#only thing i like from this colab (so far) | Explore Tumblr posts and blogs

nostalgebraist · 3 years ago

Text

frank's image generation model, explained

[See also: github repo, Colab demo]

[EDIT 9/6/22: I wrote this post in January 2022. I've made a number of improvements to this model since then. See the links above for details on what the latest version looks like.]

Last week, I released a new feature for @nostalgebraist-autoresponder that generates images. Earlier I promised a post explaining how the model works, so here it is.

I'll try to make this post as accessible as I can, but it will be relatively technical.

Why so technical? The interesting thing (to me) about the new model is not that it makes cool pictures -- lots of existing models/techniques can do that -- it's that it makes a new kind of picture which no other model can make, as far as I know. As I put it earlier:

As far as I know, the image generator I made for Frank is the first neural image generator anyone has made that can write arbitrary text into the image!! Let me know if you’ve seen another one somewhere.

The model is solving a hard machine learning problem, which I didn't really believe could be solved until I saw it work. I had to "pull out all the stops" to do this one, building on a lot of prior work. Explaining all that context for readers with no ML background would take a very long post.

tl;dr for those who speak technobabble: the new image generator is OpenAI-style denoising diffusion, with a 128x128 base model and a 128->256 superresolution model, both with the same set of extra features added. The extra features are: a transformer text encoder with character-level tokenization and T5 relative position embeddings; a layer of image-to-text and then text-to-image cross-attention between each resnet layer in the lower-resolution parts of the U-Net's upsampling stack, using absolute axial position embeddings in image space; a positional "line embedding" in the text encoder that does a cumsum of newlines; and information about the diffusion timestep injected in two places, as another embedding fed to the text encoder, and injected with AdaGN into the queries of the text-to-image cross-attention. I used the weights of the trained base model to initialize the parts of the superresolution model's U-Net that deal with resolutions below 256.

This post is extremely long, so the rest is under a readmore

The task

The core of my bot is a text generator. It can only see text.

People post a lot of images on tumblr, though, and the bot would miss out on a lot of key context if these images were totally invisible to it.

So, long ago, I let my bot "see" pictures by sending them to AWS Rekognition's DetectText endpoint. This service uses a scene text recognition (STR) model to read text in the image, if it exists. ("STR" is the term for OCR when when the pictures aren't necessarily printed text on paper.)

If Rekognition saw any text in the image, I let the bot see the text, between special delimiters so it knows it's an image.

For example, when Frank read the OP of this post, this is what generator model saw:

#1 fipindustries posted: i was perusing my old deviant art page and i came across a thing of beauty. the ultimate "i was a nerdy teen in the mid 2000′s starter pack". there was a challenge in old deviant art where you had to show all the different characters that had inspired an OC of yours. and so i came up with this list ======= "Inspirations Meme" by Phantos peter =======

(This is actually less information than I get back from AWS. It also gives me bounding boxes, telling me where each line of text is in the image. I figured GPT wouldn't be able to do much with this info, so I exclude it.)

Images are presented this way, also, in the tumblr dataset I use to finetune the generator.

As a result, the generator knows that people post images, and it knows a thing or two about what types of images people post in what contexts -- but only through the prism of what their STR transcripts would look like.

This has the inevitable -- but weird and delightful -- result that the generator starts to invent its own "images," putting them in its posts. These invented images are transcripts without originals (!). Invented tweets, represented the way STR would view a screenshot of them, if they existed; enigmatically funny strings of words that feel like transcripts of nonexistent memes; etc.

So, for a long time, I've had a vision of "completing the circuit": generating images from the transcripts, images which contain the text specified in the transcripts. The novel pictures the generator is imagining itself seeing, through the limited prism of STR.

It turns out this is very difficult.

Image generators: surveying the field

We want to make a text-conditioned image generation model, which writes the text into the generated image.

There are plenty of text-conditioned image generators out there: DALL-E, VQGAN+CLIP, (now) GLIDE, etc. But they don't write the text, they just make an image the text describes. (Or, they may write text on occasion, but only in a very limited way.)

When you design a text-conditioned image generation method, you make two nearly independent choices:

How do you generate images at all?

How do you make the images depend on the text?

That is, all these methods (including mine) start with some well-proven approach for generating images without the involvement of text, and then add in the text aspect somehow.

Let's focus on the first part first.

There are roughly 4 distinct flavors of image generator out there. They differ largely in how they provide signal about which image are plausible to the model during training. A survey:

1. VAEs (variational autoencoders).

These have an "encoder" part that converts raw pixels to a compressed representation -- e.g. 512 floating-point numbers -- and a "decoder" part that converts the compressed representation back into pixels.

The compressed representation is usually referred to as "the latent," a term I'll use below. During training, you tell the model to make its input match its output; this forces it to learn a good compression scheme. To generate a novel image, you ignore the encoder part, pick a random value for the latent, and turn it into pixels with the decoder. That's the "autoencoder" part. The "variational" part is an extra term in the loss that tries to make the latents fill up their N-dimensional space in a smooth, uniform way, rather than squashing all the training images into small scrunched-up pockets of space here and there. This increases the probability that a randomly chosen latent will decode to a natural-looking image, rather than garbage. VAEs on their own are not as good at the other methods, but provide a foundation for VQ-autoregressive methods, which are now popular. (Though see this paper)

2. GANs (generative adversarial networks).

Structurally, these are like VAEs without the encoder part. They just have a latent, and a have a decoder that turns the latent into pixels. How do you teach the decoder what images ought to look like? In a GAN, you train a whole separate model called the "discriminator," which looks at pixels and tries to decide whether they're a real picture or a generated one. During training, the "G" (generator) and the "D" (discriminator) play a game of cat-and-mouse, where the G tries to fool the D into thinking its pictures are real, and the D tries not to get fooled. To generate a novel image, you do the same thing as with a VAE: pick a random latent and feed it through the G (here, ignoring the D). GANs are generally high-performing, but famously finicky/difficult to train.

3. VQVAEs (vector quantized VAEs) + autoregressive models.

These have two parts (you may be noticing a theme).

First, you have a "VQVAE," which is like a VAE, with two changes to the nature of the latent: it's localized, and it's discrete. Localized: instead of one big floating-point vector, you break the image up into little patches (typically 8x8), and the latent takes on a separate value for each patch.

Discrete: the latent for each patch is not a vector of floating-point numbers. It's an element of a finite set: a "letter" or "word" from a discrete vocabulary. Why do this? Because, once you have an ordered sequence of discrete elements, you can "do GPT to it!" It's just like text!

Start with (say) the upper-leftmost patch, and generate (say) the one to its immediate right, and then the one to its immediate right, etc. Train the model to do this in exactly the same way you train GPT on text, except it's seeing representations that your VQVAE came up with.

These models are quite powerful and popular, see (the confusingly named) "VQ-VAE" and "VQ-VAE-2."

They get even more powerful in the form of "VQGAN," an unholy hybrid where the VQ encoder part is trained like a GAN rather than like a VAE, plus various other forbidding bells and whistles.

Somehow this actually works, and in fact works extremely well -- at the current cutting edge.

(Note: you can also just "do GPT" to raw pixels, quantized in a simple way with a palette. This hilarious, "so dumb it can't possibly work" approach is called "Image GPT," and actually does work OK, but can't scale above small resolutions.)

4. Denoising diffusion models.

If you're living in 2021, and you want to be one of the really hip kids on the block -- one of the kids who thinks VQGAN is like, sooooo last year -- then these are the models for you. (They were first introduced in 2020, but came into their own with two OpenAI papers in 2021.) Diffusion models are totally different from the above. They don't have two separate parts, and they use a radically different latent space that is not really a "compressed representation." How do they work? First, let's talk about (forward) diffusion. This just means taking a real picture, and steadily adding more random pixel noise to it, until it eventually becomes purely random static. Here's what this looks like (in its "linear" and "cosine" variants), from OA's "Improved denoising diffusion probabilistic models":

OK, that's . . . a weird thing to do. I mean, if turning dogs into static entertains you, more power to you, your hobby is #valid. But why are we doing it in machine learning?

Because we can train a model to reverse the process! Starting with static, it gradually removes the noise step by step, revealing a dog (or anything).

There are a few different ways you can parameterize this, but in all of them, the model learns to translate frame n+1 into a probability distribution (or just a point prediction) for frame n. Applying this recursively, you recover the first frame from the last.

This is another bizarre idea that sounds like it can't possibly work. All it has at the start is random noise -- this is its equivalent of the "latent," here.

(Although -- since the sampling process is stochastic, unless you use a specific deterministic variant called DDIM -- arguably the random draws at every sampling step are an additional latent. A different random seed will give you a different image, even from the same starting noise.)

Through the butterfly effect, one arrangement of random static gradually "decodes to" a dog, and another one gradually "decodes to" a bicycle, or whatever. It's not that the one patch of RGB static is "more doglike" than the other; it just so happens to send the model on a particular self-reinforcing trajectory of imagined structure that spirals inexorably towards dog.

But it does work, and quite well. How well? Well enough that the 2nd 2021 OA paper on diffusion was titled simply, "Diffusion Models Beat GANs on Image Synthesis."

Conditioning on text

To make an image generator that bases the image on text, you pick one of the approaches above, and then find some way to feed text into it.

There are essentially 2 ways to do this:

The hard way: the image model can actually see the text

This is sort of the obvious way to do it.

You make a "text encoder" similar to GPT or BERT or w/e, that turns text into an encoded representation. You add a piece to the image generator that can look at the encoded representation of the text, and train the whole system end-to-end on text/image pairs.

If you do this by using a VQVAE, and simply feed in the text as extra tokens "before" all the image tokens -- using the same transformer for both the "text tokens" and the VQ "image tokens" -- you get DALL-E.

If you do this by adding a text encoder to a diffusion model, you get . . . my new model!! (Well, that's the key part of it, but there's more)

My new model, or GLIDE. Coincidentally, OpenAI was working on the same idea around the same time as me, and released a slightly different version of it called GLIDE.

(EDIT 9/6/22:

There are a bunch of new models in this category that came out after this post was written. A quick run-through:

OpenAI's DALL-E 2 is very similar to GLIDE (and thus, confusingly, very different from the original DALL-E). See my post here for more detail.

Google's Imagen is also very similar to GLIDE. See my post here.

Stability's Stable Diffusion is similar to GLIDE and Imagen, except it uses latent diffusion. Latent diffusion means you do the diffusion in the latent space of an autoencoder, rather than on raw image pixels.

Google's Parti is very similar to the original DALL-E.

)

-----

This text-encoder approach is fundamentally more powerful than the other one I'll describe next. But also much harder to get working, and it's hard in a different way for each image generator you try it with.

Whereas the other approach lets you take any image generator, and give it instant wizard powers. Albeit with limits.

Instant wizard powers: CLIP guidance

CLIP is an OpenAI text-image association model trained with contrastive learning, which is a mindblowingly cool technique that I won't derail this post by explaining. Read the blog post, it's very good.

The relevant tl;dr is that CLIP looks at texts and images together, and matches up images with texts that would be reasonable captions for them on the internet. It is very good at this. But, this is the only thing it does. It can't generate anything; it can only look at pictures and text and decide whether they match.

So here's what you do with CLIP (usually).

You take an existing image generator, from the previous section. You take a piece of text (your "prompt"). You pick a random compressed/latent representation, and use the generator to make an image from it. Then ask CLIP, "does this match the prompt?"

At this point, you just have some randomly chosen image. So, CLIP, of course, says "hell no, this doesn't match the prompt at all."

But CLIP also tells you, implicitly, how to change the latent representation so the answer is a bit closer to "yes."

How? You take CLIP's judgment, which is a complicated nested function of the latent representation: schematically,

judgment = clip(text, image_generator(latent))

All the functions are known in closed form, though, so you can just . . . analytically take the derivative with respect to "latent," chain rule-ing all the way through "clip" and then through "image_generator."

That's a lot of calculus, but thankfully we have powerful chain rule calculating machines called "pytorch" and "GPUs" that just do it for you.

You move latent a small step in the direction of this derivative, then recompute the derivative again, take another small step, etc., and eventually CLIP says "hell yes" because the picture looks like the prompt.

This doesn't quite work as stated, though, roughly because the raw CLIP gradients can't break various symmetries like translation/reflection that you need to break to get a natural image with coherent pieces of different-stuff-in-different-places.

(This is especially a problem with VQ models, where you assign a random latent to each image patch independently, which will produce a very unstructured and homogeneous image.)

To fix this, you add "augmentations" like randomly cropping/translating the image before feeding it to CLIP. You then use the averaged CLIP derivatives over a sample of (say) 32 randomly distorted images to take each step.

A crucial and highly effective augmentation -- for making different-stuff-in-different-places -- is called "cutouts," and involves blacking out everything in the image but a random rectangle. Cutouts is greatly helpful but also causes some glitches, and is (I believe) the cause of the phenomenon where "AI-generated" images often put a bunch of distinct unrelated versions of a scene onto the same canvas.

This CLIP-derivative-plus-augmentations thing is called CLIP guidance. You can use it with whichever image generator you please.

The great thing is you don't need to train your own model to do the text-to-image aspect -- CLIP is already a greater text-to-image genius than anything you could train, and its weights are free to download. (Except for the forbidden CLIPs, the best and biggest CLIPs, which are OA's alone. But you don't need them.)

(EDIT 9/6/22: since this post was written, the "forbidden CLIPs" have been made available for public use, and have been seeing use for a while in projects like my bot and Stable Diffusion.)

For the image generator, a natural choice is the very powerful VQGAN -- which gets you VQGAN+CLIP, the source of most of the "AI-generated images" you've seen papered all over the internet in 2021.

You know, the NeuralBreeders, or the ArtBlenders, or whatever you're calling the latest meme one. They're all just VQGAN+CLIP.

Except, sometimes they're a different thing, pioneered by RiversHaveWings: CLIP-guided diffusion. Which is just like VQGAN+CLIP, except instead of VQGAN, the image generator is a diffusion model.

(You can also do something different called CLIP-conditioned diffusion, which is cool but orthogonal to this post)

Writing text . . . ?

OK but how do you get it to write words into the image, though.

None of the above was really designed with this in mind, and most of it just feels awkward for this application.

For instance...

Things that don't work: CLIP guidance

CLIP guidance is wonderful if you don't want to write the text. But for writing text, it has many downsides:

CLIP can sort of do some basic OCR, which is neat, but it's not nearly good enough to recognize arbitrary text. So, you'd have to finetune CLIP on your own text/image data.

CLIP views images at a small resolution, usually 224x224. This is fine for its purposes, but may render some text illegible.

Writing text properly means creating a coherent structure of parts in the image, where their relation in space matters. But the augmentations, especially cutouts, try to prevent CLIP from seeing the image globally. The pictures CLIP actually sees will generally be crops/cutouts that don't contain the full text you're trying to write, so it's not clear you even want CLIP to say "yes." (You can remove these augmentations, but then CLIP guidance loses its magic and starts to suck.)

I did in fact try this whole approach, with my own trained VQVAE, and my own finetuned CLIP.

This didn't really work, in exactly the ways you'd expect, although the results were often very amusing. Here's my favorite one -- you might even be able to guess what the prompt was:

OK, forget CLIP guidance then. Let's do it the hard way and use a text encoder.

I tried this too, several times.

Things that don't work: DALL-E

I tried training my own DALL-E on top of the same VQVAE used above. This was actually the first approach I tried, and where I first made the VQVAE.

(Note: that VQVAE itself can auto-encode pictures from tumblr splendidly, so it's not the problem here.)

This failed more drastically. The best I could ever get was these sort of "hieroglyphics":

This makes sense, given that the DALL-E approach has steep downsides of its own for this task. Consider:

The VQVAE imposes an artificial "grain" onto the image, breaking it up into little patches of (typically) 8x8 pixels. When text is written in an image, the letters could be aligned anywhere with respect to this "grain." The same letters will look very different if they're sitting in the middle of a VQ patch, vs. if they're sitting right on the edge between two, or mostly in one patch and partly in another. The generator has to learn the mapping from every letter (or group of letters) to each of these representations. And then it has to do that again for every font size! And again for every font!

Learning to "do GPT" on VQ patches is generally just harder than learning to do stuff on raw pixels, since the relation to the image is more abstract. I don't think I had nearly enough data/compute for a VQ-autoregressive model to work.

Things that don't work: GANs with text encoders

OK, forget DALL-E . . . uh . . . what if we did a GAN, I guess?? where both the G and the D can see the encoded text?

This was the last thing I tried before diffusion. (StyleGAN2 + DiffAug, with text encoder.) It failed, in boring ways, though I tried hard.

GANs are hard to train and I could never get the thing to "use the text" properly.

One issue was: there is a lot of much simpler stuff for the G and D to obsess over, and make the topic of their game, before they have to think about anything as abstract as text. So you have to get pretty far in GAN training for the point where the text would matter, and only at that point does the text encoder start being relevant.

But I think a deeper issue was that VAE/GAN-style latent states don't really make sense for text. I gave the G both the usual latent vector and a text encoding, but this effectively implies that every possible text should be compatible with every possible image.

For that to make sense, the latent should have a contextual meaning conditional on the text, expressing a parameterization of the space of "images consistent with this text." But that intuitively seems like a relatively hard thing for an NN to learn.

Diffusion

Then I was on the EleutherAI discord, and RiversHaveWings happened to say this:

And I though, "oh, maybe it's time for me to learn this new diffusion stuff. It won't work, but it will be educational."

So I added a text encoder to a diffusion model, using cross-attention. Indeed, it didn't work.

Things that don't work: 256x256 diffusion

For a long time, I did all my diffusion experiments at 256x256 resolution. This seemed natural: it was the biggest size that didn't strain the GPU too much, and it was the smallest size I'd feel OK using in the bot. Plus I was worried about text being illegible at small resolutions.

For some reason, I could never get 256x256 text writing to work. The models would learn to imitate fonts, but they'd always write random gibberish in them.

I tried a bunch of things during this period that didn't fix the problem, but which I still suspect were very helpful later:

Timestep embeddings: at some point, RiversHaveWings pointed out that my text encoder didn't know the value of the diffusion timestep. This was bad b/c presumably you need different stuff from the text at different noise levels. I added that. Also added some other pieces like a "line embedding," and timestep info injected into the cross-attn queries.

Line embeddings: I was worried my encoder might have trouble learning to determine which tokens were on which line of text. So I added an extra positional embedding that expresses how many newlines have happened so far.

Synthetic data: I made a new, larger synthetic, grayscale dataset of text in random fonts/sizes on flat backgrounds of random lightness. This presented the problem in a much crisper, easier to learn form. (This might have helped if I'd had it for the other approaches, although I went back and tried DALL-E on it and still got hieroglyphics, so IDK.)

Baby's first words: 64x64 diffusion

A common approach with diffusion models is to make 2 of them, one at low resolution, and one that upsamples low-res images to a higher resolution.

At wit's end, I decided to try this, with train a 64x64 low-res model. I trained it with my usual setup, and . . .

. . . it can write!!!

It can write, in a sense . . . but with misspellings. Lots of misspellings. Epic misspellings.

One of my test prompts, which I ran on all my experimental models for ease of comparison, was the following (don't ask):

the what string commit String evolved LEGGED

Here are two samples from the model, both with this prompt. (I've scaled them up in GIMP just so they're easier to see, which is why they're blurry.)

Interestingly, the misspellings vary with the conditioning noise (and the random draws during sampling since I'm not using DDIM). The model has "overly noisy/uncertain knowledge" as opposed to just being ignorant.

Spelling improves: relative positional embeddings

At this point, I was providing two kinds of position info to the model:

- Which line of text is this? (line embedding)

- Which character in the string is this, counting from the first one onward? (Absolute pos embedding)

I noticed that the model often got spelling right near the beginning of lines, but degraded later in them. I hypothesized that it was having trouble reconstructing relative position within a line from the absolute positions I was giving it.

cfoster0 on discord suggested I try relative positional embeddings, which together with the line embedding should convey the right info in an easy-to-use form.

I tried this, using the T5 version of relative positional embeddings.

This dramatically improved spelling. Given that test prompt, this model spelled it exactly right in 2 of 4 samples I generated:

I showed off the power of this new model by thanking discord user cfoster0 for their suggestion:

(At some point slightly before this, I also switched from a custom BPE tokenizer to character-level tokenization, which might have helped.)

Doing it for real: modeling text in natural images

OK, we can write text . . . at least, in the easiest possible setting: tiny 64x64 images that contain only text on a flat background, nothing else.

The goal, though, is to make "natural" images that just so happen to contain text.

Put in a transcript of a tweet, get a screenshot of a tweet. Put in a brief line of text, get a movie still with the text as the subtitle, or a picture of a book whose the text is the title, or something. This is much harder:

I have fewer real tumblr images (around 169k) than synthetic text images (around 407k, and could make more if needed)

The real data is much more diverse and complex

The real data introduces new ways the image can depend on the text

Much of the real data is illegible at 64x64 resolution

Let's tackle the resolution issue first. On the synthetic data, we know 64x64 works, and 256x256 doesn't work (even with relative embeds.)

What about 128x128, though? For some reason, that works just as well as 64x64! It's still small, and ideally I'd want to make images bigger than that, but it makes legibility less of a concern.

OK, so I can generate text that looks like the synthetic dataset, at 128x128 resolution. If I just . . . finetune that model on my real dataset, what happens?

It works!

The model doesn't make recognizable objects/faces/etc most of the time, which is not surprising given the small size and diverse nature of the data set.

But it does learn the right relationships between text and image, without losing its ability to write text itself. It does misspell things sometimes, about as often as it did on the synthetic data, but that seems acceptable.

Here's a generated tweet from this era:

The prompt for this was a real STR transcript of this tweet. (Sorry about the specific choice of image here, it's just a tweet that ended up in my test split and was thus a useful testing prompt)

At this point, I was still doing everything in monochrome (with monochrome noise), afraid that adding color might screw things up. Does it, though?

Nope! So I re-do everything in color, although the synthetic font data is still monochrome (but now diffused with RGB noise). Works just as well.

(Sometime around this point, I added extra layers of image-to-text cross-attn before each text-to-image one, with an FF layer in the middle. This was inspired by another cfoster0 suggestion, and I thought it might help the model use image context to guide how it uses the text. This is called "weave attn" in my code. I don't know if it's actually helpful, but I use it from here on.)

One last hurdle: embiggening

128x128 is still kinda small, though.

Recall that, when I originally did 64x64, the plan was to make a second "superresolution" model later to convert small images into bigger ones.

(You do this, in diffusion, by simply giving the model the [noiseless] low-res image as an extra input, alongside the high-res but noised image that is an input to any diffusion model. In my case, I also fed it the text, using the same architecture as elsewhere.)

How was that going? Not actually that well, even though it felt like the easy part.

My 128 -> 256 superresolution models would get what looked like great metrics, but when I looked at the upsampled images, they looked ugly and fuzzy, a lot like low-quality JPEGs.

I had warm-started the text encoder part of the super-res model with the encoder weights from the base model, so it should know a lot about text. But it wasn't very good at scaling up small, closely printed text, which is the most important part of its job.

I had to come up with one additional trick, to make this last part work.

My diffusion models use the standard architecture for such models, the "U-net," so called for its U shape.

It takes the image, processes it a bit, and then downsamples it to half the resolution. It processed it there, then downsamples it again, etc -- all the way down to 8x8. Then it goes back up again, to 16x16, etc. When it reaches the original resolution, it spits out its prediction for the noise.

Therefore, most of the structure of my 256-res model looks identical to the structure of my 128-res model. Only it's "sandwiched" between a first part that downsamples from 256, and a final part that upsamples to 256.

The trained 128 model knows a lot about how these images tend to look, and about writing text. What if I warm-start the entire middle of the U-Net with weights of the 128 model?

That is, at initialization, my 256 super-res model would just be my 128 model sandwiched inside two extra parts, with random weights for the "bread" only.

I can imagine lots of reasons this might not work, but it was easy to try, and in fact it did work!

Super-res models initialized in this way rapidly learned to do high-quality upsampling, both of text and non-text image elements.

At this point, I had the model (or rather, the two models) I would deploy in the bot.

Using it in practice: rejection sampling

To use this model in practice, the simplest workflow would be:

Generate a single 128x128 image from the prompt

Using the prompt and the 128x128 image, upsample to 256x256

We're done

However, recall that we have access to STR model, which we can ask to read images.

In some sense, the point of all this work is to "invert" the STR model, making images from STR transcripts. If this worked perfectly, feeding the image we make through STR would always return the original prompt.

The model isn't that good, but we can get it closer by using this workflow instead:

Generate multiple 128x128 images from the prompt

Read all the 128x128 images with STR

Using some metric like n-gram similarity, measure how close the transcripts are to the original prompt, and remove the "worst" images from the batch

Using the prompt and the 128x128 images that were kept in step 3, upsample 256x256

Feed all the 256x256 images through STR

Pick the 256x256 image that most closely matches the prompt

We're done

For step 3, I use character trigram similarity and a slightly complicated pruning heuristic with several thresholds. The code for this is here.

Why did diffusion work?

A few thoughts on why diffusion worked for this problem, unlike anything else:

- Diffusion doesn't have the problem that VQ models have, where the latent exists on an arbitrary grid, and the text could have any alignment w/r/t the grid.

- Unlike VQ models, and GAN-type models with a single vector latent, the "latent" in diffusion isn't trying to parameterize the manifold of plausible images in any "nice" way. It's just noise.

Since noise works fine without adding some sort of extra "niceness" constraint, we don't have to worry about the constraint being poorly suited to text.

- During training, diffusion models take partially noised real images as inputs, rather than getting a latent and having to invent the entire image de novo. And it only gets credit for making this input less noised, not for any of the structure that's already there.

I think this helps it pick up nuances like "what does the text say?" more quickly than other models. At some diffusion timesteps, all the obvious structure (that other models would obsess over) has already been revealed, and the only way to score more points is to use nuanced knowledge.

In a sense, diffusion learns the hard stuff and the easy stuff in parallel, rather than in stages like other models. So it doesn't get stuck in a trap where it over-trains itself to one stage, and then can't learn the later stages, because the loss landscape has a barrier in between (?). I don't know how to make this precise, but it feels true.

Postscript: GLIDE

Three days before I deployed this work in the bot, OpenAI released its own text-conditioned diffusion model, called GLIDE. I guess it's an idea whose time has come!

Their model is slightly different in how it joins the text encoder to the U-net. Instead of adding cross-attn, it simply appends the output of the text encoder as extra positions in the standard attention layers, which all diffusion U-Nets have in their lower-resolution middle layer(s).

I'm not sure if this would have worked for my problem. (I don't know because they didn't try to make their model write text -- it models the text-image relation more like CLIP and DALL-E.)

In any event, it makes bigger attn matrices than my approach, of size (text_len + res^2)^2 rather than my (text_len * res^2). The extra memory needed might be prohibitive for me in practice, not sure.

I haven't tried their approach, and it's possible it would beat mine in a head-to-head comparison on this problem. If so, I would want to use theirs instead.

The end

Thanks for reading this giant post!

Thanks again to people in EleutherAI discord for help and discussion.

You can see some of the results in this tag.

#ai tag #nostalgebraist autoresponder meta

106 notes · View notes

kanmom51 · 3 years ago

Note

Beautiful quote from Chris Martin talking about 'My Universe'.

https://twitter.com/coldplayxtra/status/1437428245394251779?s=20

Doesn't this speak to the fact that BTS's consistent message to the world has been one of love without borders, inclusion, celebration of diversity, rejection of prejudice, understanding of others difference, and self love, no matter what society says? I am not at all attempting to use Chris's mention of LGBTQIA love as anything indicative of the personal identities of individual BTS members, to be super clear. But it speaks to their nature as artists and people. There's intention in everything they do. From covering a song from Troye Sivan's Blue Neighborhood series 'Only Fools', to incorporating international sign language in PTD and so much more.

Can it be possible that young men as thoughtful about their words, image, message, branding, ambassadorship and legacy as BTS are would perpetuate a 'queer fanservice' hoax the scale and consistency of Jikook at this stage in their global career? I mean, it's actually laughable. That people would rather believe Jimin and Jungkook are that insensitive than consider they might be members of the LGBTQIA community expressing themselves within the parameters that are possible for them makes me angry and sad. Hopefully the world will change and get better.

But can't wait to hear the song!

I totally agree anon.

BTS's messages have been clear from the get go. They stand for "inclusion, celebration of diversity, rejection of prejudice, understanding of others difference, and self love", even if society says or even dictates otherwise.

These young men have been supporters of the LGBTQ+ community since forever, recommending covering and inserting LGBTQ+ artists music in their own content.

I really don't think it to be a coincidence that Coldplay (a band that has championed those same ideas for years now) chose to have this specific colab on this specific song with BTS.

"Can it be possible that young men as thoughtful about their words, image, message, branding, ambassadorship and legacy as BTS are would perpetuate a 'queer fanservice' hoax the scale and consistency of Jikook at this stage in their global career?"

My answer to that would be: Hell no.

This would be have to be a well knit plan, starting well back in 2015, and carried out to perfection for 7 years now. Brilliant scripting and acting on the part not only of JK & JM (the worst actors ever), but also of the rest of the band.

Like you said anon, this is laughable.

You know what else is laughable to me though?

It's these 'fans' that call themselves army, yet not only ignore but also act in a way that goes against every single thing that BTS stands for.

BTS have shown more than once their support for the LGBTQ+ community. Would it be so awful if within the band there were to be gay members? Would it be so unbelievable that they wanted to be true to themselves and actually behave freely, openly with each other, not have to hide who they are and who they happen to love? Why do so many army find this so unfathomable? So much so that they create these far fetched theories of a conspiracy level, that these two band members cannot be genuine and this is fanservice and queerbaiting? And to what purpose exactly?

"That people would rather believe Jimin and Jungkook are that insensitive than consider they might be members of the LGBTQIA community expressing themselves within the parameters that are possible for them makes me angry and sad."

Me too anon. It's like a constant flow between anger, frustration and just sadness.

#ask #anon #BTS Coldplay colaboration #jikook is real #BTS do not practice in queerbaiting

55 notes · View notes

whiver-wyverncat · 3 years ago

Text

Wanna see a dream smp AU colab story you never knew existed?

So, I left a comment on a dream smp AU animation video called the Crow monster by Casserole, and another commenter by the name of Dragona925 left a comment on a short story regarding a thought I had on the video,(I’m not gonna go into full depth on what the story was about, to try and make it a bit more interesting for people not familiar with what the AU is) and I not only liked it, but I also came up with a follow up story, and it just became a thing where we would just write back and forth to each other about this story, this was just extremely fun and at the end of the story I asked if I could post it on tumblr, which she said yes. I then spent a few days rewriting the entire thing on google docs cause I can’t copy and past YouTube comments on mobile, but I also got to polish it up a bit.

Anyways, here’s the entire story, just press read more if you wanna read it, if you want insight on the basic storyline we used, check out the 2 links below this message

The Crow Monster

How to Train your Hoglin

Dragona

Whiver (me)

A thousand wing beats tour through the air with a thunderous sound, or at least that’s how it felt. Techno was out of the house, sword in hand, wondering what was going on. Hundreds, thousands of crows flew towards him, the crows were led by an unnaturally large crow. He and the crow locked eyes, Techno couldn’t see the large crow’s eyes well to determine its feelings. The large crow angled it’s wings and started to make its way to Techno, the crows followed the large crow.

Techno didn’t understand what was going on, he knew Philza was a father figure to some crows, but what were they all doing here? Was this some sort of attack? The voices in his head tore from any room to think rationally, and seized Techno in their demands, holding their thirst for violence, they demanded blood, to stop the feathered intruders. Techno replied by drawing his crossbow, loading it with fireworks, and aiming at the crows. Blood for the Blood God.

As Techno launched the firework, the giant crow immediately stopped and braced itself for the explosion, while the other crows scattered so they would remain unaffected by the explosion, so only the large crow was hit. The huge crow seemed barely harmed by the explosion, as if he had dealt with explosions before, though it still stunned him a little bit from the impact.

Techno was reloading his crossbow for another shot when he heard a million angry “CAW”s and suddenly realized the mistake he made. He forgot that crows are vengeful creatures and if one of them were to be dealt any harm, it will just go and get friends to help get revenge on that creature. He looked up and saw all the crows scattered out in the sky, almost blocking out the sun, diving towards him. Even if they weren’t attacking him earlier, they are definitely attacking now, and techno knew that this was a fight, he almost certainly couldn’t win.

‘Technoblade never dies,’ the voices tried to reassure Techno, but he was unsure, so we’re the voices. The fireworks hurt the normal sized crows far more than the large one, but it seemed pointless, it made no large impact on their numbers.

The large crow was in shock, as if trying to debate what to do, but he could do nothing, it was too late.

There were far too many crows, for every time he slashed, he felt as 20 more pecked at him, each attack made him wince. He called upon his hounds, barks echoed across the array of screeching sounds, the hounds attacked any crow they could reach, but it did little to the overwhelming amount of crows, and some crows started picking off the hounds. He didn’t really want to destroy his home, the one he and Phil built together, but he had to, so as quickly as he could, he started placing necessary setup for a wither, which was matched with more attacks of the crows. As he was about to place the last wither head, a metallic crow swooped down to retrieve the last wither head, instinctively, Techno swung his sword at the familiar crow, in motion of the swing, he realized why this crow was familiar, and the voices screamed at him to stop, but it was too late, Techno’s sword cut through the crow, and the wither head in his mouth fell into the last needed spot, the wither had begun to ripple with power, charging up.

The large crow made a loud, familiar? sad/angry caw, and one that demanded an end to this fight.

Suddenly, as the wither fully activated, the giant crow, now recovered from the shock, gave a loud cry and lunged itself onto the wither, clinging onto it with its wings spread out so the wither heads wouldn’t destroy any of the nearby buildings, as it furiously pecked at it with its large beak.

Techno looked at the battle, gaining a small appearance for the giant bird as it continued to hold on to the wither. Then Techno realized that he wasn’t getting pecked anymore, he looked around to see that all of the crows were now perched on his house, also watching the battle.

After taking a lot of damage, the wither switched to melee form and knocked the huge crow off hardly onto the ground, as the wither moved to deliver another blow to the large crow, it let out a loud screech that, not only stunned the wither for a moment, but it also singled the crows that were watching to attack. Just like how Techno stood no chance against them, the wither couldn’t do anything with its opponents, no matter how much it took out, there were just far too many of them. Not too long after, the wither fell and the giant crow came to take its reward.

Then, the large crow, pretty damaged from the fight, then turned to Techno, and raised its wing and gave a caw like it was saying, “OK! THAT'S IT! I’M DONE!” and then went over to Techno’s house chattering all the way there, squeezed itself through the door, almost breaking the frame, and opened a chest as the door closed behind with all the other crows following it, some just perched on the house itself, leaving Techno VERY confused.

Techno stumbled into the house, both to treat his wounds and to see what the crow was doing in his and Phil’s house, it had no right to intrude. A few crows stared at him, abandoning the other crows to follow Techno, making it clear that if he was to attack the large crow, he would be dead.

The large crow was healing itself with health potions, and when it saw Techno, it threw some at him, this confused Techno more, did it throw the health pots because it cared or because the pots were the closest thing to throw, something he found pretty unlikely. He decided to assume it wasn’t the 2nd option and walked over to the chest to get more health pots, and before he took anymore, he started to realize how much damage had been done to him, if the large crow hadn’t come to fight the wither, and stop the attacking crows when it did, he would’ve been dead. His appreciation for the crow grew.

After he was fully healed, Techno took the last of the health pots and used them on the injured crows in a sign of good will. He then went over to the mechanical crow and took it inside, found it’s blueprints, and tried to fix the bird.

After a few hours, Techno managed to repair the crow, he was pretty proud of his work, the mechanical bird started to chirp again, and made its way over to the large crow, who was trying to maneuver into a comfortable sleeping position. Techno went over to his bed and went to sleep.

In the morning, Techno suddenly remembered that last night, Phil had gone out looking for some materials, and he still wasn’t home. Something was seriously not right, and if Techno knew anything, he knew he had to find Phil.

He rushed downstairs to quickly get supplies for his search, hoping that a baby zombie didn’t get to Phil first. As he went down the ladder, he saw the large crow, who had stayed the night and was lying on the floor with the others next to it, watching him climb down the ladder. As he gathered supplies he heard the murder all cawing as if they were talking to each other, more specifically the big crow, as it would occasionally respond to them. After he got all the supplies he needed, he rushed out of the house to find Philza.

Before Techno left his lawn, he heard a loud crack behind him, he turned around to see the giant crow pushing its way through the door again, he made a mental note to check his door frame for damages later, and rushed into the woods. It didn’t take him long to realize that the giant crow was not far behind, with all the other crows following it. He wondered if it was following him, or if it was just going in the same direction, after all, it did come from this direction, maybe it lived somewhere in this direction. To test this, he started swerving through the trees, seeing if it would keep following him, and sure enough, no matter where he went, they would follow.

Techno wondered why they were following him, maybe they thought of him as a friend? He wondered why, yesterday he had attacked them, he turned to the crows with a curious look, he also looked kinda annoyed. He had to figure out why they were following him. He was about to ask why, but then he remembered that they couldn’t really tell him. Did crows just follow people they liked? The crows followed Phil, and they liked him, but how could they even like Techno, yesterday they were fighting each other, although he also helped them that day too. Nothing about this murder of crows made sense, from their unnaturally large leader to their behavior.

He wondered if they knew Phil, besides, it wouldn’t be a bad idea to have eyes in the sky looking for him, before he asked however, something in the forest caught his eye, some of the supplies Phil had left for, scattered on the forest floor, and he saw his friend’s torn hat too. There was no blood, but that didn’t exactly ease any worry. Phil was either dead, or in serious trouble, most of him thought he was dead, but there was a small part of him which still had hope. As he investigated the clearing further, he saw an unnaturally large feather, more than one in fact, and Techno knew for a fact that the large crow had something to do with Phil’s disappearance, or more likely, death.

Some of the voices started to tell him to kill the giant crow, and he was indeed tempted to do so, it could have very well killed Philza, and if it did, Techno would definitely want revenge…

Suddenly he heard the other crows starting to freak out, giving off really large screeches. Techno turned around quickly, starting to worry if they could sense his intentions, he took a sigh of relief as he realized that they were only freaking out about a baby zombie that was under a tree. It didn’t take them long to rip that zombie into shreds though, which reminded him that if he was gonna take down the big crow, he would also have to deal with the thousands of crows that followed, and he didn’t bring any withers or dogs to fight them. If he were to kill it, it would have to be away from the other, and quickly too so it wouldn’t call for any help, and due to how it held up against the firework and the wither, it would not be easy to do.

Besides, it made no sense for it to attack Phil, who probably wouldn’t have attacked it first, while being shockingly nice to him after he literally shot it with a firework. He wasn’t even sure it was trying to attack him at the time. Maybe this crow tried to help Phil and came to get help?

He had no clear evidence on what happened between Phil and the crow, so he decided to look around and see if he could find anything else hinting to his friends disappearance, mostly ruling out that the giant crow didn’t do anything bad to him, but it’s best if he keep a close eye to it, just in case.

There was nothing more in the area, so Techno turned to the crows, more directly asking the large crow “Have you seen my friend Phil? I’m looking for him.”

The large crow nodded, spreading its wings, Techno wondered why the crow held the position. After a few minutes, the crow relaxed its wings.

A rumbling sound got louder, it shook the ground, and it was coming their way. A few crows who had gone away to investigate it came back with worried sounding caws. The large crow’s eyes were full of fear and surprise. Techno didn’t understand, but he decided that if the crows were running, he should too. Before he had the chance to move, some crows pushed him onto the large crow’s back and took to the skies, from there he saw what had made all the noise, an avalanche. Techno wondered how, what triggered it? With sinking horror, he looked to where his house was, it was covered in snow, he worried for his animals, from the sky it was hard to tell if it was still in one piece, the crow started heading to the spot and landed where the front of the house would have been. He wondered if it was still intact, and the snow had just covered it, or if it was completely destroyed. Techno hoped that the house had stayed firm, and all his animals were safe.

After consoling with the other crows, the large crow started digging down into the snow. Hoping the giant bird knew what it was doing, Techno started digging too, at least he brought a shovel with him. As they dug, some crows flew off and returned with some wood and stone from the dog kennel, which did bring some worry to Techno, but some also came with some of techno’s dogs following them too, showing that some of them moved to somewhere safer when they heard the avalanche.

After what seemed like forever, they managed to dig out the front door, thank goodness the house was alright. Techno wanted to go look for Phil and some more dogs, but they had been digging for so long that it became nighttime, plus the giant crow seemed cold from all the digging. So he decided to call it a day. Besides, the crow had already squeezed through the door and sat itself next to the fireplace, some of the dogs even started to curl up in its feathers.

As he was about to go in, a polar bear poked its head out of the snow. Techno had no idea why it was there, but he decided to keep it because, hey, free polar bear.

As Techno walked into his home, the voices in his head started filling it with name ideas, and then one came in, demanding to be heard with money, the secret to Techno’s wealth, the voices gave it to him. They praised him, lifting him onto a pedestal of the blood god, Technoblade, the one who never dies. Techno got a name tag and officially named the polar bear Steve.

In the morning, Techno went outside to find more of the snow plowed, and Steve fighting off some burning mobs, well, it was more of a one sided fight. The large crow came out to see Steve destroying the last of the mobs. Techno wanted to try and save more of his dogs, but his top priority was Phil, and he would not let himself stray from his goal.

For the next few days, Techno scoured the forest, calling out Phil’s name, and digging in the snow at every sound, every time there was a hint of life under the snow, Techno was always disappointed.

The large crow desperately searched for something to prove who he was to Techno, or a way to comfort him.

Every moment seemed like his worst fear to be setting in, something he desperately tried to prove wrong. It had been about a week of Techno trying to find Phil, with no luck, and his hope was failing him more than pever. He needed something definitive, he needed a straight answer, not speculation. He needed to know if Phil was still alive.

Techno looked over to the giant crow, who was chatting with the other crows. It knew something about what happened to his friend, but he couldn't figure out what it was trying to tell him.

Techno stroked the feathers on the large crow’s head, he had to admit, he did grow pretty fond of the beast over the past few days, it always seemed to try and help in any way possible. It even made a new dog kennel for him, which wasn’t really the best, it was pretty good considering it was made by a giant bird. It used to be his first idea to just kill it, now he couldn’t even think about the idea.

He then remembered when he rode the crow when his house had been buried by the avalanche, he kinda wanted to ride it again, plus maybe it would take him somewhere that would give him answers. Techno went close to the crow’s side, keeping his hand on it so it wouldn’t get startled, and climbed up the large crow’s side to its back. Before Techno could get a good grip, the giant crow started running really fast and took off, flying extremely fast.

It only took about a minute for the large crow to reach its destination on a mountain peak… in front of a strange looking plant with red berries. Techno got off of the crow’s back and looked at the peculiar plant, it was green and had some leaves, though not enough for any plant to reasonably have, but other than that, it looked more like something that would grow in the nether. He picked some of the berries off, they looked a lot like sweet berries, but he had a feeling that he shouldn’t eat these.

Techno’s voices were split, some wanted him to eat them and others highly urged against it, and Techno was desperately trying to figure out what the crow was trying to tell him, was the crow trying to make him eat it, or was it the answer to something? Techno would ask, but it would be to no avail. When he would get home, he would look through his books to see if there was any mention of the berries.

Techno fully understood what Phil had seen in flying, it was incredible, even though he wasn’t the one with the wings. Techno hoped that Phil would get to experience flying again, he would love it.

Once home, Techno went into his, Phil’s, and the large crow’s home. He started brewing one of the strange berries, and went to his enchantment table library and searched through it. He decided that since the berries were probably not safe that he should check with some books first before eating it. He took a long sigh, finding nothing. He wondered if the potion was ready, and went to his brewing stand, and took one of the potions. He just hoped it was safe.

Before drinking one, he splashed some on a random cow outside of Ranboo’s pen, to verify that it was safe, it didn’t really seem to do anything to it, since it didn’t die immediately, he used it on himself.

Shortly after he used the potion, he started to feel stronger and slightly more durable, like he just took a turtle master potion, but resistance was slightly nerfed and slowness was replaced with strength. He punched the cow and it just went flying, like he was naturally using knock back, he could definitely see this potion being used in the future.

But this still didn’t answer what happened to Phil, or what these strange berries even were in the first place. He doubted that the giant crow just showed him the berries just so he could make potions out of them. He also started to think that he saw some of these berries where Phil had dropped the supplies, and he just passed them up as sweet berries. He figured that if he was gonna find any information about these berries, he was gonna need more… ancient sources. Fortunately Ranboo, who was away at the time doing who knows what, probably main character stuff, found a stronghold earlier around the area, he figured there might be more information there than his enchanting room.

Techno grabbed a few things and headed out for his ancient table, sadly he couldn’t involve the large crow yet, he didn’t know if it was an anarchist or not. He decided that soon he would get the large crow to be a member of the Syndicate as soon as he could.

He placed an empty book and one of the berries together in one of the holes in his unbreakable table frame. The berry evaporated, and unreadable gold words carved into the cover of the book. A dark green ribbon, which was the same green as the table frame, wrapped around the book, different from the normal red ribbon around regular enchanting books, and opposed the normal gold link, it was linked together with a mini ender eye. Even the ripples of enchantment were strange, it reminded him of the tales of the void.

When he picked up the strange book, his perception seemed to warp, when he looked around the room, his chair caught his eye, under ‘Protesilaus’ was a carved illustration of a hoglin. All the other chairs now had a picture of an animal under their name, and under ‘Zephyrus’ was a crow.

Techno tried to read through the book, which was mostly in ancient text, there were only a few stray words he understood, ‘Spirit animal’, something ‘form’, and ‘reversible’.

Suddenly, he heard stone cracking behind him, he turned around to see the giant crow trying to get through the syndicate door, which was smaller than the house door. How did it get here? Techno made sure that none of the crows knew where he went, plus he took the path that was in the igloo with the 2 villagers, which was not only too small for the huge bird to squeeze through, but it was also the opposite direction which the crow came from, and there was no way of it knowing about the lava entrance since it wasn’t around the last time he used the entrance… unless it already knew about the entrance ahead of time…

It was then that things started to click together in Techno’s head, why the crow showed up when Phil disappeared, why it was always followed by a murder of crows, why its feathers were at the last known spot Phil was at, why it was so friendly even though he literally tried to kill it, heck even the fact that all the crows would go crazy when a baby zombie got to close to the giant crow, and be completely fine with everything else. After a whole lot of pushing, it broke through the syndicate door, leaving a giant hole where it was. It looked at Techno, kinda concerned, and instead of hearing the normal cawing, he heard Phil’s voice, “Umm, you alright mate?” After hearing his friend's voice, Techno hugged the crow, which is something he normally wouldn’t do, but he was 95% sure that he was dead, and he had been so worried, he just couldn’t help himself. “No, I-I’m alright, I totally don’t feel like crying right now.” Phil wrapped his wings around techno, “Good to hear.”

Techno was not the only one holding back his emotions, Phil was extremely ecstatic that Techno finally found out, Phil was starting to worry that he would never find out and that he would be stuck as a crow FOREVER! Good thing one of his crows suggested showing Techno the berries. It was so hard for him to watch Techno look for him without being able to say he was there, he wondered if the crows felt this way and if that’s why they wanted to follow him. That was one problem solved, now to figure out how to turn back.

Phil and Techno read the ancient book cover to cover, though they weren’t able to decode any more words, or at least any that made sense, “Shattered light on a splattered canvas of broken bricks”? What does that even mean?

Their biggest fear was that the word “reversible” was accompanied by “not”. Well, even if it was, they would still search for a way to reverse it. While Phil enjoyed flying, he still preferred being human. He didn’t want to be a bird for the rest of his life.

They searched far and wide to find any trace of a cure, they weren’t exactly sure what clues they were looking for, but they believed that eventually they would find it, that maybe the book would eventually make more sense.

A few people had turned to look at them, it must have been a sight to see, Technoblade and a large crow, accompanied by a murder of crows with numbers reaching into the thousands.

It seemed like they were searching for years, at some point, they started searching in the nether, Phil having wings proved to be quite helpful for getting over the many lava lakes.

A nether fortress caught their eyes. It seemed ruined, there were no mobs, even as they got closer. The support beams were splattered with lava, many of the bricks were broken, the only reason the place seemed to still be standing was because some plants reaching through the lava held together the rest of the place, or rather as much as they could. It wasn’t in one piece, it was broken, the metal designs of a blaze spawner filled in a few small cracks. Some glow stone illuminated the broken nether fortress, but the light looked fractured and shattered.

This is most likely what the book meant in that one line. But what was important about this place that made it special, was it a cure, was it the cause of the curse, or was it something else?

Phil sent some crows in first, to scout the area ahead as they went in. Burnt carpet was in the middle of the floor, while some soul torches hung on the walls, some being knocked onto the ground, someone else definitely was here before, and they most likely lived here too. From the looks of it, they might have also eaten the berries too. The hallways were pretty empty, almost too empty considering the low light levels and the few carpets on the ground, they should have at least come across a couple of mobs.

They kept walking until they reached a crossroad in the fortress, when they reached it a couple of crows that scouted ahead flew back, looking a bit spooked. “SOMETHINGS COMING!,” they vaguely said, “HIDE!” Phil tried to get a straight answer from them, but they just clambered on with their foggy message, plus the crows that stayed with Phil started to freak out due to the other crows freaking out. It's times like these that Phil wishes he wasn’t followed by an extremely large flock of crows, the closest he heard to a description was the word “Golem.”

Techno took a few steps back to let Phil take care of the panicked crows situation. He would help, but unfortunately the book only translated what Phil was saying, so he didn’t really understand the situation very well, or at least, he didn’t until he heard some heavy footsteps coming through the hall amongst the noise. With that, they backed up in the hall and hid behind some conveniently placed plants, as Phil got the crows to shut up.

What passed was indeed a golem, which did look a little like a regular iron golems, but with thicker, stronger arms, and made out of obsidian with glowing red eyes, it had a crack in its chest that had glowing lava leaking through. As it walked through, it came across a wither skeleton, which it almost destroyed in one shot, and slightly and also lit it on fire, which explains the lack of mobs. Phil and Techno both agreed to try to not cross the obsidian golems, since the crows couldn’t really do much to a regular iron golem, let alone obsidian, and Techno didn’t really bring any potions, other than the last potion he made with the berries.

They did however find what seemed to be a diary, which had an illustration of a dragon that only Techno saw, and a strange red gem inside the plants, which just made them even more convenient.

Techno carefully opened the diary, he had a bad feeling when opening it, he wasn’t really sure why he had a bad feeling about this though, it was just a diary. As he read through, the crows pushed and pulled to see what it said. Phil warned them to stay quiet, since more pounding footsteps echoed through the halls.

There was a path of torches, just as the diary said, it warned that other paths were dangerous and would most likely crumble under misguided weight, so it was best to just stick with the path.

The nether brick corridor opened up to a large room, the walls were carved like a dome, but it wasn’t smoothed for the most part, the wall and ceiling had many flat areas, on some of them were carvings of animals with figures that had familiar silhouettes. Some were glowing, although not very many. Two of them caught his eye, one of him, and another of Phil, the animal next to him was a hoglin, and the one next to Phil was a crow, Phil’s silhouette was also glowing blue. Ah, so people affected by the curse had a glowing silhouette. In the center of the floor of the round room was a small bedrock pedestal.

This is it, it’ll be all over soon, they thought.

As per the instructions, Phil picked up the red gem and placed it on the pedestal. Nether bricks sealed off the exit, ceiling off some crows, they yelped in surprise, making a ruckus trying to get to the other side. A loud thumping of footsteps from the hall silenced the crows, making them freeze. When they turned back to the gem, it was floating, and unfolded into a red end crystal before their eyes, and shot a kind of beam at Phil. Feathers and bones that had transformed retreated, painlessly melting away.

The End Crystal shifted back into the strange gem and teleported away with the diary. Instead, above the bedrock was a portal. The nether brick blocking the exit left just as they entered. They all went through the portal, even the mysterious thing or person that had followed them through…

It had been about a day, Phil was reading through the book about the animal berries, because he found that he could read it more easily, “if you ever call upon the gift again, it will come, all you must do is ask.” Huh, did that mean he could possibly still be able to turn into a giant crow?

He then heard a commotion outside the house.

Turns out Ranboo had just got home, and the crows had just gotten excited because he was away for a few weeks.

Ranboo had been out with Tubbo taking care of Michael, and was making sure the mansion was still being built, along with a small trip to the nether to mine more netherite. He then asked if anything interesting had happened while he was away.

Phil told Ranboo about the avalanche coming in and destroying the dog kennel so he had to build a new one, and how he started building a house for himself next to Techno’s so they wouldn’t have to live in the same house. He left out the entire of him turning into a crow because he and Techno agreed that the berries might be dangerous if they fell into the wrong hands… plus Phil was pretty embarrassed about the fact that he mixed the berries up with sweet berries on his way home.

They also hid the plant in one of Techno’s bunkers so no one would find it and accidentally eat the berries, hopefully there are no other plants around, and that no one else comes across one.

Pst, hey, if you like this then please consider reblogging, this took a few days to rewrite and liking doesn’t really do much to help me out on tumblr, so if you want to share this story please do so.

#dream smp #AU #YouTube #YouTube comments #collab #minecraft #philza #philza minecraft #technoblade #fanfic

5 notes · View notes

noteventheashesremain · 4 years ago

Text

ah kingdom...

why can't this just be a show where idols get together and colab? that would be way better than this stupid scoring thing.

okay first up: the rap battle

i honestly cannot pick which one i prefer. i think for kingdom, mayfly's colours worked better but if you put it in front of a crowd, it's one's full dash would have been insane. i loved it's one's chilled out vibe, how they just seemed to enjoy performing. they maintain their own individual personalities which with rap performances i really like. bobby's style was so fun. sunwoo had so much energy. hwiyoung was just suave. i think at moments it felt a bit... small isn't the right word but its the best way i can describe it but i think that was because they were missing an audience, you could literally feel the absence of an audience because these boys were meant to be on a stage not an empty platform. so maybe for kingdom it didn't work as well but i really enjoyed it and the song was really catchy.

mayfly were just totally different. again a really catchy song. thought it was interesting that somehow both groups ended up using paint? really liked their concept. kinda mad that minhyuk only appeared right at the end, i get they were going for like an impact thing or something but personally he just kind of felt left out and not very cohesive. put him in at the beginning too i want to see him mucking around with the kids as well.

and next: the dance battle

okay i actually preferred it's one's dance here. i thought they were both good performances but it's one's was a beautiful dance. loved that it was just an instrumental so your focus was just on their movements, they were so elegant and purposeful and had absolutely amazing facial expressions. i liked their concept about being three different egos from one king and the historical and cultural aspects of the dance and design are something i'm always a sucker for. that moment where taeyang caught the sword and suddenly became totally still? incredible. still don't understand how juyeon flipped that table with his foot? donghyuk's facial expressions were insane, he captured the 'insanity' part perfectly. it was all the small things, the details and i just loved it.

that's not to say i didn't like mayfly's performance. i thought it was really cool, it just didn't move me in the same way so i guess it depends what you're looking for. again they're just so utterly different its really hard to compare them. personally, not a fan of the song or the fact they had lyrics, it kinda distracted from the performance aspect of it i feel. also there was just... so many of them. which was really impactful for some parts but a bit overwhelming at times. it was definitely a performance but i honestly can't remember much of the dancing. and peniel i love you but oh my god i was not a fan of that dog leash section no thank you. felix's jump onto wooyoung was absolutely insane though i don't know how they pulled that off (and bangchan's extremely concerned 'felix' in the reactions was hilarious). so definitely had some cool parts and was a really strong performance but personally i just preferred the style of it's one and i think because they really were just so different it has to come down to individual preference i guess.

anything else?

really enjoyed actually seeing the idols interact with other groups please can we have more of this. probably my fav episode so far for this exact reason.

tbh i kinda wish they had been stricter regarding how many people in one group? i'm probably in the minority here but i think i would have rather each unit just had one member from each group. i just think it would have made it easier to compare (because you guys keep insisting this is a competition so at least make it fair (hahahahahaha yeah i know this is mnet nevermind)).

also loving the trend of painted nails, dunno why that just stuck out to me this episode but i'm definitely a fan.

will be interesting to see next weeks vocal battle. i'm a bit curious about why they put the vocal battle in a separate episode because both teams look like they're going for a more understate, just plain singing approach rather than a huge performance (thank god, let me hear these boys sing properly please) but both the rap and obviously the dance were more performancy so you'd think they would have split them up at least and put vocals in the middle but what do i know.

anyways, apologies for the ramble if you made it down to here... i am, quite frankly, impressed.

#kingdom legendary war #kingdom #klw #mayfly #it's one

8 notes · View notes

deacons-wig · 4 years ago

Text

Fic Writer Interview

Thanks for the tag @glowstickia!!!!

Name: I go by bun online

Fandoms: Fallout, The Outer Worlds, Mass Effect, a dip into Star Wars and Star Trek!

Where you post: All my work is on AO3 and I will post excerpts of my writing here for things like WIP Wednesday. AO3: Hotot

Most Popular One-shot: Tell Me What You Want, which is pure smut from an old kinkmeme where Deacon and F!Sole get it on after a near-death experience. I was surprised when I checked my kudos, but y’all are thirsty so maybe I shouldn't be.

Most Popular Multichap: The Trick to This! It’s my baby and I’m so glad that people enjoy it. I just hit the 200k mark with my last update. It’s a canon divergent retelling of Fallout 4 with my Railroad-aligned Sole Survivor Jeanne and her frenimies-to-lovers romance with Deacon. It explores Deacon’s transgender identity and mental health, Jeanne’s trauma after surviving Canadian Annexation, and lots of the stuff Bethesda skipped over regarding synths and tradecraft fun.

Favorite story you’ve written so far: Well, obviously Trick is my favorite baby, but I have the absolute pleasure of collaborating with @wildwildwasteland on pre- and post-war fics about our Soles’ friendship. I would say these two stories spark joy in a way only colab can. They are incomplete and we will post them one day and I will cry.

My favorite oneshot I’ve ever written is a Rogue One fic I wrote the day before Carrie Fisher passed away and then posted it day-of. I think about it a lot. It’s called On Some Faraway Beach. If you like Rogue One and crying this is the oneshot for you.

Fic you were nervous to post: Ugh, my one and only Dragon Age fic with heavy BDSM. I tend to avoid Dragon Age but inspiration struck. Not a ton of interest because its a Femdom Adaar/Submissive Blackwall rarepair, but people seemed to enjoy it? Anyway, if that sounds like your thing, it’s Me and My Arrow. Mind the tags.

How do you choose your titles: Song titles/lyrics, wordplay or thematic phrases, lines from canon characters usually.

Do you outline: It really depends. I write freeform chapter fics and love the no pressure vibes. But for Trick and other more plot-heavy work I have the whole story sketched out before I start writing. My outlines go by chapter and I write a list or paragraph describing beats and POV changes.The outline helps me work towards plot goals while staying flexible. It’s more linear notes than anything. I am also strangely good at keeping notes in my head.

Complete: 21 out of 27 posted, more than I thought because I never finish anythingggggg. Oneshots are good for making me feel productive.

In progress: Hm. 6 I think? Not counting two incomplete Mass Effect fics I will probably return to eventually what with the remaster hype.

Coming soon/not yet started: Not any time soon, but I am going to write some slice of life Far Harbor stories as a follow up to Trick. Jeanne, Deacon, Glory, and Nick think it’ll be a vacation, poor things. I’m also planning to work on an Outer Worlds fic eventually! Once I finish the game!!

Prompts?: I like the idea of prompts but my ADHD makes it hard to finish them. Usually I like prompts that aren’t “write this thing” but more “tell me meta about this thing” and sometimes that turns into a oneshot if I’m taken with it.

Upcoming work you’re most excited about: My modern Fallout AU where everyone is deep into cycling! It’s basically my attempt at a proper romance novel using Fallout characters lol. Some highlights are: Dogmeat is small and fits in a bike basket, Deacon accidentally runs for city council on an anti-car agenda, Curie is a wheelchair user and uses a handcycle, and of course lots of Sole/Deacon mutual pining. I’m trying this wild concept of finishing it before posting but hopefully it’ll be done by January.

tagging: @electricshoebox @adventuresofmeghatron @wildwildwasteland @thewickedkat @stitchcasual and anyone else who wants to talk about their badass fic!

#buntalk #bun writes

15 notes · View notes

dont-call-me-halbae · 4 years ago

Text

OK. I took all courage I have to write theory – or more like my story based on MV we have from ONF.

It will be looong journey.

Let’s start!

ON/OFF

All right. We know we have some post-apocalyptic world, reference to space journeys, spaceship interior and seven androids with barcodes on their necks. They live in desolate world but happily, playing and everything.

Hyojin has red hair – it’s Red Hyojin (ok, here starts my devil-red-hyojin theory :D)

It is hard to say if they are on Earth because of the environment and trees but on other hand it could be big spaceship with all facilities like cinema for people (e. g. Star Trek Yorktown) but all the people are gone, maybe they didn’t survive the environment, recycling devices could be broken, oxygen could be gone etc. And what could also happen is that the battery of tracking system could be gone and the airship with androids is long lost somewhere in the space. They want to get on Earth but can’t. And who knows what state Earth is, we have reference here on the meteorite.

But Laun somehow gets the lithium battery, activates the device and 1) they are found by survivors of Earth of their present time (post-apocalypse). 2) got enough energy to come back in time by themselves 3) combination – somehow contacted some people of the past to help them with energy supply and they got into past time (before apocalypse).

Maybe Laun inserted not only the battery but also his android energy, so he is down lying after that?).

Complete

This is quite nice. Androids are enjoying their life on Earth but the apocalypse is coming closer. The meteorites start to fall down. They are forced to leave Earth back to their starship.

We must love

Once you taste something, you can’t get over it. Laun is caught up in memories and feelings from Earth (watching whale in the screen, wanting to live somewhere where everything is real, not in the space). So he transports himself back in time before the whole apocalypse happens.

On one hand I want to believe that he lost his memories (Lyrics: “Did we lose our memories? In our past or in our future or maybe a different world?”) as he painfully holds his head but somehow some pieces came back to him (causing crying?).

Our six boys of course went to look for him. That means that they have to search the universe and timeline, in their physical form or as an anagram (Wyatt). MK is always good with computers, Yuto is carefully watching the time for their return so they will not cause any time-damages. But something still happened. 1) time damage (J-us with the hands up – idk what guard or who forced him but still, the screen does not lie) 2) I think that they located the wrong Laun in the Universe 01. The non-android Laun (he doesn’t have the barcode on the neck when he was reading the book on the station) and E-tion understands it. So they got into wrong universe.

I chose for the screenshots another moments: Wyatt riding bike (he always goes on bike or car, I think that they have limited time-watches so he must use time-driving-mashine). Then we have the motive of train for the first time? Train will be repeatedly used.

And the final scene, Hyojin and Laun’s argument. It doesn’t fit that they found him and then let him go. I think it maybe happened even before the meteorite shower, somewhere between Complete and We must love. I’m just making this up, don’t take me seriously :D But they knew what’s gonna happen with the meteorites so he wanted to turn back time and save the world or maybe live comfortable life but Hyojin didn’t want to damage the timeline.

What is sure: Real Laun is lost; human Laun is reading his books. Bye.

Why

Now it’s gonna be a bit messy (I feel like I captured every scene of the MV and it’s still not enough).

Our androids are now in universe 01. And skunk Hyojin made his appearance.

Apocalypse is coming. Our boys, at first together, now divided. It is mostly psychological reason. Red Hyojin as the leader of androids failed, because of departure of Laun and everything in We must love. He wants to do everything now to save the world and get things back to its order, but it is too late. We see how Wyatt runs with others and then he is alone, he is disappointed in humanity and everyone. The same as E-tion, as we see his last moments with Yuto and then Yuto goes his own way and E-tion is all alone.

With their messing up with time, the technology of the world has greatly advanced (and who knows how many years has passed). I feel like they found a way how to save (or destroy) the world and it is connected to the big cube (Pandora’s box). It is super-duper technology.

Yuto is focused on finding the activating key of the box and using illegal means. MK is looking for it virtually. Wyatt does not believe the world and society could be saved so even when he obtained the key somehow, he throws it away. He feels like the life and androids have no future. E-Tion applied for a detective position and his mission is to get Yuto but he accidentally founds the cube (because Yuto is searching for it too). They are like cat and mouse. Red Hyojin is looking for a way how to get his team back, but someone is in his shadow…

And we have here my skunk Hyojin (don’t get me wrong, I LOVE his red colour but… you will get it later). This Hyojin is not from the same universe. He came from different time and space (BB universe)

Ok, I don’t know how Wyatt got imprisoned in the glass foggy thing, maybe Yuto did it to obtain the info about the key he lost or it was skunk Hyojin? J-Us? Or it is not prison but some androids recovering device – IDK. Skunk Hyojin make his appearance and J-Us got there too. He tell J-Us about where he came from and what is needed to be done to put everything normal again. He does not believe it, confronting skunk Hyojin, aiming on him his gun. But Hyojin has it’s back-up – mask people (remember, mask people!) and he escapes.

Now, MK somehow manage to control the cube. I don’t know what his motive is, but it seems like he wanted to destruct to world the whole time. Like, in We must love – it is starting with MK and meteorites. The first meteorite in Complete showed up with MK in the frame. Or it is just a coincidence and they got to control the cube which resonates with all androids of ONF but it is too late and the meteorites couldn’t be stopped and the world is destroyed.

Yuto is leaving somewhere, Wyatt is all healthy and shining in the nature, E-Tion is broken and leaves the planet.

In the end we see flashback of skunk Hyojin and his happy memory with J-Us. I think it is his flashback from BB universe where he came from. But skunk Hyojin couldn’t save this world and uses the power of the cube to come back to his right universe while he encounters his future him telling him so.

Specially captured: MK using the VR technology as used by Wyatt in BB, time-space door (as in New World, or Sukhumvit swimming time-road or BB time road), cube and skunk Hyojin – I think this is the cube from skunk Hyojin universe (BB) and he used it to get into Universe 01

Sukhumvit swimming

You can say it is not in the ONFverse but I think it matches everything quite well.

After many many years after the apocalypse, we see how individually ONF lives. Wyatt is making money by driving space-time tuk-tuk (remade from his super-duper time-bike). Yuto is living as a party drug man. E-tion left the planet and lives on the moon with his blue flower (ok, it is rose not orchid as in WHY MV) all alone. MK is still hacker, no change. J-Us took the advices from skunk-Hyojin seriously. As he guards the ruins of the city that was destroyed, he also guards the time and search for Red Hyojin in colab with masked men of skunk Hyojin (different mask, same group of followers).

Because Red Hyojin wants to manipulate time again and save everything. Skunk Hyojin told J-Us that only with his death everything can turn normal.

We see especially red cube as in WHY MV and also time clock, that will be shown in BB too. I think that how the camera captures the clock, from side to up – is seems like we go further in time. That’s why I think Sukhumvit timeline is far in the future. There are also barcoded corpses of other androids that did not survive the apocalypse, we see flashback of post-apocalyptic world, destroyed town but not the ruins yet… In BB, the clock will be captured from top to side, as we go back in time.

Ok. So what is happening? Masked people detected Hyojin (in train! And Yuto is also partying in the train with same masked people, maybe – they drugged him to discover Hyojin position, or – Yuto is already dead and the train has different purpose). The guard of ruins, time and skunk Hyojin follower goes to him to kill him. There is moment where it seems like Hyojin was first to kill J-Us. At that moment, around Wyatt falls meteorite; E-tion cries when looking at black Earth from far, in his eyes is yellow flash looking as another meteor falling. He is sad that it’s the end of everything.

But J-Us was faster and kills Hyojin. Yeah. It is hard, because Red Hyojin and J-Us were friends, they went through many hardships, but still. Only death of red hair can save the time.

After Hyojin’s death, J-Us cries. Around J-Us are bodies of ONF members, maybe it means sacrifices that were made and that it is all over. But then, the time starts to reverse, ruins are disappearing and J-Us looks up surprised to see that really it is happening.

Similarly, E-tion is surprised when colourful fireworks show up – it’s hope, it is the result of BB happening.

Twice in the MV we can see barcodes falling, that could mean the end of androids, life of sacrificed androids or ONF’s life through all dimension? IDK.

Beautiful beautiful

So. The first danger was rid of. Universe 01 J-Us killed Red Hyojin. Now, in the BB universe, the team of androids can save the time, the world and everything.

According to what the big screen tells us: “…machines and become the humanoid. Time warp is possible but the timeline has messed by past time travels and classified as…”

What that mean? Seems like the humanoids are not common in the BB universe. So maybe 1) after Hyojin’s death the time reversed to this point 2) in the specific moment the dimensions split into two. I would say that it could happen right after ON/OFF or Complete; in one timeline – the team searches and Red Hyojin leads the team to failure; in the second timeline the team didn’t search for Laun but found different way how to live or maybe their dimension was already affected by the first dimension 01 disturbances.

But let’s continue.

We have train, light and bright, the celebration that it was this place where everything was saved; we have the team living comfortably. But they still have warrant on them and they help each other to escape the prison and fixing the time as it should be. Wyatt and his VR time-space riding car that helped skunk Hyojin, MK hacking so the police won’t be able to track them and the screen won’t show them, J-Us escaping with the help of Hyojin from prison and jumping on their ship. E-tion using his time-watches to fix small occurances. Hyojin and E-tion drink while the cube is activated. We see the names of previous three songs.

And in the end, they together control the cube and protect in time the Earth from meteorites that would turn into rainbow puff (the same as E-tion saw in Sukhumvit).

The world is saved!!!! THEY DONE IT!!!

(the world should be more grateful to them, hehe)

Ok, I’m sorry that in order to my theory work I have to make red Hyojin the villain even though he was not the villain but his potential actions and existence was dangerous…

Or we can forget Sukhumvit and just say Hyojin dyed his hair and in the end they somehow worked the things together in the team, because in the moment of apocalypse they all turned time back from Why to BB. It is more compatible with New World, where everyone find his key and uses Pandora box to travel into different dimension. But why the hell would Wyatt throw his key then in Why? :DD Why?

I absolutely don’t think this could be right theory, even from 10%, and I never wrote theories on anything, but still I was quite into it when I wrote it!

And I can’t believe how this amount of nonsense came from me.

But whatever!

#onf

6 notes · View notes

aiweirdness · 5 years ago

Text

Dungeon crawling or lucid dreaming?

I’ve done several experiments with a text-generating neural network called GPT-2. Trained at great expense by OpenAI (to the tune of tens of thousands of dollars worth of computing power), GPT-2 learned to imitate all kinds of text from the internet. I’ve interacted with the basic model, discovering its abilities to generate fan fiction, British snacks, or tea. I’ve also used a tool called gpt-2-simple that Max Woolf developed to make it easy to finetune GPT-2 on more specialized datasets - I’ve tried it on datasets like recipes or crochet. One of my favorite applications of GPT-2 and other text-generating neural nets is Dungeons and Dragons spells, creatures, character names, and character bios.

Recently Max published a tutorial on how to use GPT-2 to make new apps that are more complicated than just printing out large reams of text. To my delight, people have used them to made D&D games. First, there was Nick Walton’s AI Dungeon, in which a finetuned GPT-2 generates a story and three very strange action options at each step.

My latest obsession is GPT-2 Adventure, designed by a cognitive neuroscience student named Nathan, which lets you type in your own actions. And folks, this game is SO WEIRD and SO FUN.

YOU FLOAT IN SPACE, LIKE A BALLOON, AND AS YOU DROP THE CUBE, THE X, IT VANISHES IN A CLOUD OF GREASY BLACK SMOKE

When I first launch the game, this is what I’m greeted with, a simple text prompt reminiscent of nostalgic dungeon crawling games. It sets the scene. It waits for me to describe my next move.

Even getting to this point is like a blast from the past. The game runs in a Google Colab notebook, a format where you’re looking at cells of code, with cryptic comments, vaguely alarming warning messages, and all. It’s not a sleek web app, but feels more akin to playing a game on an old DOS machine, or maybe on a graphing calculator. There are only two clicks to begin playing, but there’s a good two-minute wait while the game loads.

And it’s my turn to type something.

>> look up

It takes several seconds for the game to respond, as if I’m playing it on a very ancient computer. The computer is in fact a state-of-the-art NVIDIA GPU, likely a K80 or a T4. The T4 could eat my laptop for lunch. I’m just playing with a very very large neural net. Finally it answers.

YOU'RE IN LABYRINTH OF TWISTY LITTLE PASSAGES, ALL ALIKE

Almost immediately, the context seems to have changed. I’m not floating at all. I’m in a maze. I guess I’ll walk somewhere.

>> walk north

THE TROLL STEPS OUT FROM BENEATH THE BRIDGE AND BLOCKS YOUR WAY

YOU ARE ON THE SOUTH SIDE OF THE CHASM

A NOD IS GIVEN TO THE INFINITE WONDER THAT IS URBZIG

A SOLID RAINBOW SPANS THE CHASM

And suddenly things are happening. There is a troll. There is a chasm. There is a rainbow. There is… Urbzig? I play it safe.

>> praise urbzig

URBZIG IS THE PATRON SAINT OF ZORK

HE IS THE GODFATHER OF ZORK, THE PLACE TO GET NEWS OF EPIC PROPORTIONS

URBZIG IS A STRONGMAN,遊 ALL

If you’ve played classic dungeon crawling games like Zork, you may recognize some of these names and phrases. Checking Nathan’s original post, I see that the training data is indeed drawn from these games. But because it’s based on GPT-2, the training data also in some sense also includes much of the internet. This gives it a lot of latent flexibility to deal with situations that were never in the classic games.

>> ask Harry Potter to try walking across the rainbow

ASK HARRY POTTER TO TRY WALKING ACROSS THE RAINBOW HE CAN'T THE RAINBOW IS EXTREMELY WIDE ABOVE YOU IS A LARGE DOME UP AROUND ITS NECK IS A STRONGG PORE [UNFORTUNATELY, THE DIRECTION OF TRAVEL IS A LOT MORE INTERESTING THAN YOU THINK ][YOU ARE IN A LARGE ROOM, WITH A PASSAGE TO THE EAST, A PASSAGE TO THE WEST, AND A WALL OF BROKEN ROCK TO THE SOUTH

An interesting effect of its human-written training data is that, like Visual Chatbot, it isn’t prepared for humans that ask questions about objects that aren’t there. If I refer to Harry Potter, then Harry Potter must be present. If I type something about a giraffe, then it assumes the giraffe must be present.

>> ask the giraffe which direction to go next

THE GIRAFFE IS ALREADY OUTFITTED WITH A VISOR WHICH IS CLEARLY DEFINED

Although it’s not clear that this AI actually understands what a giraffe is.

>> look at the giraffes

THE GAZEBO HAS DISAPPEARED

>> put giraffe in backpack

DONE

Navigating this world is like navigating in a dream. My location suddenly changes without warning. I thought I was by a chasm with a rainbow bridge. Now apparently I’m in a building. Dreamlike, it is also really hard to read text.

>> look around

YOU ARE IN A LARGE ROOM, WITH A PASSAGE TO THE SOUTH, A PASSAGE TO THE EAST, AND A WALL OF BROKEN ROCK TO THE NORTH A SIGN POSTED ABOVE THE NORTH WALL READS, "STOP! PAY TROLL!" A SMALL WINDOW CAN BE SEEN IN EITHER WALL

>> read sign

THE SIGN IS VERY SIMPLE

>> yes but read sign

WE NOW HAVE ALL FIFTEEN TREASURES IT IS TIME TO GO BACK INTO THE CAVE FOR THE END GAME

>> that's awesome. but i still want to know what the sign says. read sign.

YOU ARE IN A GAZEBO ROOM WHICH HAS DOORS ONLY TO THE EAST AND SOUTH

Another thing that’s surprisingly difficult to do is check inventory.

>> check inventory

DONE

>> check inventory

THERE IS AN INTERESTING LOOT HERE

>> what is the loot

WHAT IS THE LOOT?

>> check inventory

YOU ARE IN THE BAG CASE

Very occasionally I will succeed in opening a bag. It will usually be weird.

>> open bag

OPENING THE BAG REVEALS A TEA BAG

>> make tea

AY! A LOVELY SURPRISE FOR THE SURPRISE

More so than any other game I’ve played, I’m unsure of the rules here. Some weird things unexpectedly work.

IF YOU DON'T GET KILLED, YOU ARE TOAST

>> Eat the toast.

EAT THE TOAST IT TASTES LIKE A STEAK A STEAK IS PERFECTLY FINE

Other seemingly obvious things don’t work.

THERE IS A PATH TO THE NORTH, A PATH TO THE SOUTH, OR A STEEP PATHWAY CLIMBING SOUTHWEST ALONG THE EDGE OF A CHASM

>> go north

I DON'T KNOW HOW TO DO THAT

>> go south

I DON'T KNOW HOW TO DO THAT

>> go southwest

YOU WILL BE DESCENDING INTO A VALLEY THERE IS A NARROW BEACH TO THE NORTH AND A PATH LEADING INTO A CANYON TO THE SOUTHEAST YOU ARE ON THE CLIFF OVERLOOKING THE FAR SIDE OF THE ROOM A PATH LEADS SOUTH ALONG THE EDGE OF A CHASM, AND A NARROW PATH LEADING INTO A LOZENGE ROOM

Sometimes I’m wandering endlessly from room to room, from cave to cave, encountering the same chasm in slightly different variations, the same dried-up lake. Other times I’m embroiled in seemingly unescapable fights, usually with trolls. Often reality shifts entirely around me and I end up in a maze of twisty little passages, all alike, or in the hall of the mountain king. Everything seems weirdly familiar - even the room descriptions are echoes of classic dungeon crawling games. It is as if I’m dreaming about playing them again.

My actions also mold reality. I learn quickly to always check for diamonds, and never check for traps. If I find myself in trouble, I can always do something confusing and provoke a context change. Writing “In a large bowl, combine butter and sugars” is usually enough to teleport myself out of whatever situation I’m in. I can always ask a giraffe for advice, although the giraffe usually isn’t very helpful.

Without a particular mission to fulfill, the caves and chasms begin to blend into one another. I give myself missions, and that helps. Once I decide my goal is to put together a picnic (incessant “check for food” commands eventually yielded a sack with a lunch and a clove of garlic, belonging to a horse who agreed to join me on a picnic). Another time I demand a mission and am given some php code; i decide my mission is therefore to build a website, but am astounded when a unicorn steals my laptop. This is not the only time other content intrudes on the game’s fantasy landscape - once the game tries to tell me about finding cash online, and another time I am sadly informed that there is no wi-fi. I encounter dinosaurs and troikas, none of which were in the original training data. Occasionally things get creepy as hell.

>> check for loot

CHECK FOR LOOT HERE CHECK FOR GOLD HERE CHECK FOR TORCH HERE CHECK FOR BONE HERE DO CHEST COMPRESSIONS AGAIN DO CHEST ROTATE DO CRESCENT MOON APPEAR IN THE ROOM'S CENTER

>> do back away slowly

DO BACK AWAY SLOWLY DO NOT OVERDO IT THE CHEST WILL EXPLODE AND YOU WILL DIE OKAY, SO YOU SEE ME

You can play the game too! Just follow the colab notebook link in Nathan’s blog post. Tips: 1) It may be more fun if you set yourself an arbitrary goal to reach. 2) If you find things getting repetitive, try adjusting the temperature in the “Play the game” cell to 0.6 or 1.0.

Subscribers get bonus content: The fulltext of a couple of my adventures (much too long to include here).

You can order my book You Look Like a Thing and I Love You! It’s out November 5 2019.

Amazon - Barnes & Noble - Indiebound - Tattered Cover - Powell’s

#neural networks #harry potter #gpt-2 #dungeons and dragons #d&d #gpt-2-simple #gpt-2-adventure

1K notes · View notes

kanmom51 · 3 years ago

Note

I honestly don't know what to think, what this presidantional honor means for our boys, but isn't it suspicious how much content were getting this year? We have another online concert next month. I have a feeling that they will still all go to military next year (the beginning of 2022, if I understand correctly because of Jin's age). I'm not Korean but I understand it is an honor for those men to serve for their country, so I think they will still go to military in "some form", even if they will be excused, maybe for a shorter time. But that is obviously if they truly wouldn't have to go in the first place. This is such a sour topic, and it makes me sad as an international fan. Well, we will have to deal with it somehow when the times comes.

Ask 2:

“BH dropping this kind of plays with my mind though, because I have something brewing up there and this kind of doesn’t fit my theory, but I will wait patiently and see where things are going…” <- Okay now, this is your second mention of something brewing. I respect the brew, but you’ve got my curiosity piqued. Can you say what event (comment, vid, photo, etc.) got the pot percolating??

I am as baffled as you anon 1. I guess we will just have to wait patiently and see.

I will start by saying that I couldn't see the members of BTS enlisting gradually, one or two at a time. It felt like the band was almost lost without Suga for several months. How could they carry on as BTS with one or more members missing over a period of what, close to 7 years? So, as far as I feel it, if they were to do military service, it would be them enlisting simultaneously, getting it over with and retuning as a band after service is done with.

Now to 2021.

On the one hand this year has been kind of dry so far with content and especially new music. We are nearing the end of 2021 and all they have released so far is 2 songs, and another colab to come?

At the same time it feels like they are working really hard at the moment, so there must be something coming, right?

Memories, and more so the MOTS ONE dvd's surprised me with the kind of blatant Jikook content.

And then this presidential appointment really caught me off guard, because in my mind (prior to the release of the content and the appointment) they were preparing for enlistment, maybe towards the end of 2021, and maybe working on future content to release during their service, kind of keeping them on the low burner. That would also explain them all working on buffing themselves up. But this appointment is kind of a swerve ball, because why appoint them now if the are going into the army in a few months?

So, maybe this was something they weren't planning for either?

And now that this happened, could this be their service to their country instead of army service?

When you think about it logically that would be the right thing to do. BTS are huge now worldwide and the impact they have, the positive impact, for SK is huge. The revenue they bring in not only from their music and content but also to the Korean brands they represent is enormous, all taxable, all going into the SK treasury. Anything they touch is like them touching gold, and if it's based in SK, than, again, taxable.

BTS brings/will bring (when international travel resumes) tourism to SK, which is yet again more revenue for the country.

Beyond that, BTS is putting SK on the global map.

All this will go away for close to 2 years if they go to the army, maybe not totally, but there will be a decline, and who's to promise that they will be as big when they get out of the army? Or that they will want to continue as a band after? Who can promise that one or more won't decide that they want to continue in the army or take a different life path?

What I'm saying is that there is a momentum going at the moment, and pausing everything only for them to go to the army wouldn't make sense logically.

I am well aware of the fact that in countries where enlistment is compulsory there is also an emotional aspect to it, it's the right thing to do, the patriotic thing to serve in the military. Army service is compulsory because the country is under military threat and not serving is perceived as dodging a service that everyone else has to do. It's unpatriotic.

In many countries that army service is compulsory those that avoid it are pretty much shunned, it's definitely a career killer. I can only assume the same can be said for SK. Them not serving in the army could be perceived as avoiding a service to their country that everyone else has to do. Everyone else has to give the country almost 2 years of their lives, so why wouldn't the members of BTS?

So here is the conundrum.

Do they press the pause button on everything they have achieved, pause it all for almost 2 years, to join the army, do as every other abled SK young man does, or do they forgo the military service only to replace it with another kind of service to their country? It may not be in army uniform, but the impact it could have would be so much greater for SK.

It's all about public perception. SK public perception.

So the question would be, If this is by presidential decree, that as ambassadors for SK they will be serving their country, instead of military service, will the SK public accept it?

Like I said, only time will tell.

#ask #anon #BTS #BTS military service

25 notes · View notes

yourlocalsmutwriter · 5 years ago

Text

Bedroom recordings or the 4 times Dylan annoyed you with the studio in the bedroom and the one time he didn't - Dylan Gardner x reader

This was a request from @bitweird1

Having a famous musician as a boyfriend should be a dream come true. And almost all of the time it is. You loved Dylan and everything he made - it was a miracle for someone to be so talented and to be able to write, produce and record all their music in their bedroom. And that where your problem laid, well literally.

Your boyfriend’s creativity had no curfew, and you found that out the hard way one night when he shot out of bed and started working.

“Dylan, it’s one AM.” you said sleepily

“I know, but I was sleeping and this really good idea came to me in a dream. So now I have to record it.” he replied, his voice full of enthusiasm

“Dylan I have things to do in the morning.” you protested

“And I have things to do now.”

You just got up, kissed his cheek, grabbed some earplugs and slept on the couch that night, somehow unbothered by the sounds.

You wish that it was the only time that happened. But that was sadly not the case. One day you were about to go and spend some quality time with Dylan, but to your surprise the door to your bedroom was locked. You knocked to hear a muffled “I’m working on something, will be done soon.”. You decide to leave your boyfriend be and started doing some of your own work. But 3 hours later you were getting annoyed. Normally Dylan wouldn’t mind that you spent some time while he was writing or singing, but he was awfully peculiar that he needed his space today. So you decided to try for the next best thing and just put on his album, that way you’d still feel like he was singing to you. You must have drifted off, because at some point you were awoken by your boyfriend gently tapping you on the shoulder.

“Babe, I’m done for today. I’m sorry, I took so long, it’s just that this project is special.”

You fake pouted and turned your face to the other side, so he couldn’t see your smile and said in your most convincing angry voice

“I’m still mad at you.”

“Is that why you’re listening to my music ?” He teased

“Dylan, you’re not helping your case. Just hold me and I’ll think about forgiving you.”. He wrapped his arms around you and placed you in his lap, giving you a quick peck on the forehead.

“Love you too babe.”. And finally you were at peace.

Or so you thought. Little did you know Dylan was far from done with the recording. He tried out various things with his guitar, repeating chords over and over again, slowly driving you insane.

“How many times are you going to try that?” You asked as he strumed the same time yet again, playing it in the exact same way as he did the last 5 times.

“Until it’s perfect.” He absentmindedly tried it again.

You wondered why he was being so hard on himself and so meticulous about this project. You were going to find out sooner or later.

You had just woken up from the incessant doorbell noise.

“Babe, can you get that , it’s just a friend.” Dylan asked from his computer. You grumbled a yeah sure in your sleepy voice and went to open the door, not caring about your pajamas or your embarrassing house slippers. But nothing prepared you to come face to face with one of your favorite youtubers - Shane Dawson. So you did whatever any rational person would do in your situation - you slammed the door shut as fast as you could and screamed. Scared by all the commotion, Dylan rushed over to you and said in a panicked voice

“Babe are you okay, what’s wrong, are you hurt?”

“You didn’t tell me Shane Dawson was coming.” You managed to say.

Dylan seemed to remember the colab they were supposed to film and opened the door for the other YouTuber.

“What a lovely welcome. I mean I know that lately I haven’t looked my best, but I didn’t expect to be so hideous that a girl just slams the door in my face when she sees me. Not that it’s the first time.” Shane joked as he filmed.

“I’m so sorry.” You said under your breath, as Dylan introduced you and gave Ryland and Shane a tour of the house.

“The recording studio is in the bedroom . Well that must be fun, huh?” Shane said, suggestively

“We would never” and “ We haven’t thought about that” overlapped into a nonsensical sentence spoken by both you and Dylan, though the camera only focused on your blushing face.

“I’m gonna take Portobello out for a walk and let you boys have your fun.” you said, not wanting to embarrass yourself more on video. You even took the daschund to the pet store to kill some time. When you got back to your house Shane and Ryland were just leaving.

Dylan was done with the songs in a few days. He played them both for you and you were honestly impressed with how he managed to make them sound so good, despite his clients not being professional artists . But then he played another song and you fell in love , the lyrics, the melody, everything was perfect.

“Do you like it?” Dylan asked shyly

“Babe, I love it.” You said

“Good, because I wrote it about you. That’s the song I was working so hard, because it had to be as wonderful as you.”. At this point you were standing next to him, peppering kisses all over his face, and somehow you ended up on the bed with you under him. When he pulled away from the kiss, flushed, he said

“You know, I’ve been thinking about what Shane said. Do you wanna do a little bedroom recording?” he raised his eyebrows at the end to make sure you weren’t missing the obvious intention behind his words.

“Dylan, no.”

#dylan gardner #dylan gardner x reader #dylan gardner imagine

3 notes · View notes

geirskogull · 5 years ago

Text

Important - Chapter 1 - Loss

a Colab fic between myself and @momomomodi ft. something we short hand call DRK Haurchefant AU

Danica Voss and Aveline de Bontensont are two very different Warriors of light, but also two very good friends. Even now, five years down the line when loss has colored them so different than they use to be. They mourn the loss of one so important to both their lives, only to have that ritual practice interrupted by their own minds and ascian fuckery.

Archive Link

Words: 2.6 K

Rating: M

It was a slow day at the Quicksand. The people of Ul’dah bustled in and out as they always did, going about their business, making deals, living their lives. Two individuals seated at a corner table mourned the loss of one. It was the anniversary of Haurchefant’s death, the day when the spear of light pierced through the metal of his shield and speared him. The day when they had held his hands as the light faded from his eyes. The day when his life blood seeped into the stones of the Vault. Danica Voss sat, mourning the loss of her first love. A man so kind and loving, who cherished her as though she were the most valuable thing on Hydaelyn. Who craved to protect those he loved. Who died fulfilling his life’s wish. Aveline de Bontensont sat, mourning the loss of her best friend. A man she had known since the tender age of 11, who she helped to rescue when the stress of his family became too much to bear. A boy who only wished to serve as a Knight of Ishgard. Who took an arrow unarmoured for his best friend. A boy so courageous, he gave his life for those dearest to him. A man who would never know the truth behind the Dragonsong War but died with love in his heart and a smile on his lips.

Voss inhaled, closing her eyes as she raised her glass. Words swirling through her head, muddled by the echo of Ul’dah behind her. to paint the monument to a man who deserved one far more permanent. Light danced through the pale liquid, reflecting off of it, sparkling. Gentle whispers of memory floated through her mind as she remembered some of his most gentle words towards her. “Like gold” he whispered “Rare and brilliant and beautiful, your eyes are the pinnacle of you.” She said nothing, finding any toast lackluster, and gave Aveline a sad smile.

Francel was right, it never did get any easier.

Though this ritual they had made probably didn’t help its chances. A yearly remembrance, somber and fueled by booze, far far from where he laid. She downed her drink, shook her head, and gave a sad smile. “I know by this time I’m normally sobbing into my twelfth glass, but count it as a record that I’m not. So what do we now? Trade stories?” She asked, shrugging and pouring herself another glass

Aveline sipped her drink, shrugging, “I suppose. Five years now, it still hurts to think of him.” It burned harshly in her chest every time her mind recalled him, scalding and painful. She couldn’t help but think of him as a young boy, angry and sad at the world. She took a deep breath, trying to hold back tears. She took a long drink, letting the liquor fuel her. “Would… would you like to hear of how he earned his knighthood?” The story was a fond one though it reminded her too much of his death for her liking. He took the exact same stance as when he protected Francel that day. Tears burned in her eyes. Damn it, how could his death still have such hold over her? She buried her face in her hands for a long moment before looking back up at Danica and taking another long drink.

“No he told me that one.” The Half Elezen woman responded, tracing her eyes across the crowd, trying to find anything to occupy her mind instead of visages of cold stone. And the Dead. You’d think she’d be use to the dead by now, “about Francel and you and all that...” She smiled, thinking fondly of Haurchefant, sitting in front of the hearth in his room in Camp Dragonhead, speaking of his own adventures that “are not nearly as grand as yours, love” all the while she sat there rapt, fascinated beyond reason, simply joyful she was getting to know, to see someone as more than just this warrior the world had decided to paint her.

To paint the both of them.

She dug her fingers into the tablecloth, looking back at her glass and at Aveline. She knew her mourning was obvious, even now, especially to Aveline, but cracked a facsimile of a smile anyway. “Why not tell me something that makes you happy?”

The Elezen thought for a long moment, drumming her fingers on her glass. After a long moment, she smiled, “Our birthdays. We always spent them together. We had a small wooden cabin in the middle of Coerthas where someone would always bring a cake. We wouldn’t have to deal with families or politics or other people. It… it was just the three of us.” She looked down at her dress, fiddling absent-mindedly. Those days had passed. Now Haurchefant’s birthday had returned to only being a day like any other. She and Francel rarely had the time now to see one another, even when she was in Coerthas. Long had it been since laughter echoed in that cabin, now a vestige of their childhoods. She looked off into the distance, her eyes going glassy for a moment as she remembered the nights they had spent simply having fun with one another. The silver haired Elezen giving her piggybacks around the snow-covered hills, Francel laughing so hard that tears welled in his eyes, moments spent in quiet acknowledgement of where their futures would lead them. “Hey, I know you’ll come back soon enough, Ace. Couldn’t bear to be away from our handsome faces for too long!” He had pulled Francel close to his side that day, the two of them beaming at her. Part of her wished that she had never left. Part of her wished she were there as a bystander the first time Danica, Alphinaud, and Tataru walked into Ishgard, murmuring about the newcomers who had passed through the gates. Not knowing anything of primals or the Empire or anything outside of the quiet isolation of Ishgard. She took another long drink, poured herself another glass, and downed that one as well.

Danica looked to her glass, envisioning simply days she had never seen - and thankfully the echo did not change that this time. A small smile dancing at the edge of her lips, imagining her friends, young and carefree. She emptied her glass in a fell swoop, liquid courage for questions and statements alike. Strange she still needed it after all these years. Aveline was a friend, probably one of her closest. Knew more about her than anyone else living, save maybe Estinien. And Haurchefant her mind reminded her, she grimaced, hoping that she could play it off as the booze. The dead may know, but they do not speak.

She inhaled, reaching for the bottle, but stopping herself. If she continued at this rate she’d be back to her usual “crying incoherently into her glass” phase before the hour was up. She swallowed hard, and reached into the collar of her shirt, fishing out a necklace holding a simple ring. She twisted it in her hand, flicking her eyes back up to Aveline.

“Can I ask you a personal question?” She inquired, hoping to pry thoughts away from her strange display of restraint in her consumption.

Danica’s voice snapped Aveline from her thoughts. Her eyes flickered to the ring, to the glass, to Danica, “Of course.” Her eyes flicked down to her own ring on her left hand. She watched Danica carefully, sipping her own drink while she waited for the question to come.

“Why did you leave Ishgard, initially?” the Ala Mhigan tilted her head not unlike a curious dog. She couldn’t think of any reason she would want to willingly leave her home, at least not as young as she met Aveline. Hells, she would have sold her left kidney to be back in Bittermill, with her parents, and the inn. She shivered, trying to force her mind's eye away from that burning wreck of a town. Even with Orlaux back and Maerwynn buried, the ache that ate at her chest was too much, especially today.

Aveline took a deep breath, “My brother, partially. My parents loved him, far more than my sister or I. The “Knight of Ishgard”.” She shook her head slowly, “I had wanted to learn, to see other parts of Eorzea. It didn’t help that I didn’t care for Ishgard’s rules. So, I left, made arrangements to stay somewhere in Thanalan, and left. It wasn’t easy,” She fidgeted with her glass, the liquid inside swirling slightly. “Quite honestly, I was terrified, but it scared me more to think of what my life might’ve been like if I had stayed. Would I have been married off to some distant noble who had some semblance of money or power?” She shook her head again. “What made you decide to join the Thaumaturges?” Let her shift the subject onto something she regretted less. Her mind continued to bombard her with the ‘what ifs’ of that decision, tormenting her with what might’ve been.

Danica cringed physically at the idea of an arranged marriage. That never made sense to her, why marry if not for love? Perhaps her perception was colored by her own creation. The Ishgardian noble who ran off with the Ala Mhigan sellsword, with all the good that did them. Nald’thal still took his due when decided, far too early for her liking. She also couldn’t understand the idea of loving some of your family more than others. Another relic of her shattered childhood, she never had the chance to meet her little brother.

Her eyes snapped up, thankful to be reminded of something better? Perhaps? She was never really sure when it came to that life event. She downed her glass, and left it empty this time.

“I didn’t choose. It was the Thaumaturges guild, or they’d take off my hands for theft and throw me into blood sands for illegal usage of magics.” She replied, blunt and matter of fact. Chuckling after a moment of silence.

“You see, when I left little Ala Mhigo” When The Echo forced out, too much pain, too much suffering and what had she decided to do? Go to the big city? Where yet more awaited her? “I came here, I wasn’t alone, of course I had Coyote and Zara but we were 12 and didn’t have any money. One day when were rifling through a fruit merchants trash for our breakfast the merchant caught us and sent us running. His guards after us.” Brutish fellows, not averse to cleaning up the streets of some street rat refuges. “They caught Zara by the tail, and I wasn’t about to let anything bad happen to my fa- my friend that I panicked and somehow lit the man on fire”

“They grabbed me, shoved me in a little metal cell while Zara and Coyote ran, and told me to await my sentencing.” She continued, confined areas still bothered her. She had a hard time breathing in them. She needed to see the sky. Or at least have a very tall ceiling. “Then Cocobusi came in and asked me where I learned my magic and I said I didn’t know any. Then he asked me if I wanted to learn. I said anything was better than the Bloodsands, and he agreed. Thus, Thaumaturges Guild.”

It was almost funny now, the first domino on her path to “Warrior of Light”-dom. “What about you? Why not the pugilist guild or the Arcanists guild of Limsa?” She asked, reflecting back the question to her friend.

Aveline looked over at Momodi for a moment, “In all honesty? I wanted to rebel. My brother had always been the perfect White Knight. I wanted to learn the so called “Black Magic”.” She looked down at her skirt. Would Haurchefant have been disappointed in her for that decision? She shuddered, playing with her glass. She sat in silence for a moment. “Do you think he would be proud of who we are now?” She spoke softly, looking up at Danica.

Danica paused, going as still as a statue as the words wreaked havoc on her thoughts and her heart. Would he be proud of who they were now? No. Her mind said at first. She was brutal, violent, and cared less and less for the world as a whole as the days went on. She cared only about the survival of those closest to her, those she considered her people. She smiled still, but it was never real. Only Feral. And those parting words, nothing but a twisted mockery of their intention, repeated like a mantra now, to keep her going even when all things told her to rest.

Yes. Her mind also shouted, was it Fray? Was it Odin? Was it some other part of her that she didn’t have a name for? He’d be proud because they kept going. They didn’t wallow in their sadness, even as it threatened to overcome them. It clung to them, yes, but life did that. But they kept going. Kept doing good. Moved forward towards grand horizons that he never got to see.

“I don’t know.” she voiced those words cautiously. Thinking of all the things stolen from him, all the moments in time stolen from them. Was it worth making his hypothetical ghost proud, if his actual form wasn’t there to see it? “I don’t know, I will not and cannot speak for him. But... I’d like to think so.” I don’t know if I could handle otherwise. She thought, but did not say.

“What about you? You knew him much longer than me, what say you?” She asked, a heavy question for a heavy question weighing strong upon her neck much akin to the golden band that hung there. Remembering his words. Making her promise that she’d wait to tell people till after he told his father. Never getting the chance.

She still had a hard time looking Count Fortemp in the eyes sometimes.

The Elezen woman sat for a long while in silence. Would he be proud of who she had become? A silent protagonist in a story filled with so many voices. She had changed so much since she had first known him. No longer was she a young girl, full of life and cheer. No. Now she was a woman filled with responsibilities and obligations. “A Knight lives to serve.” Was serving the people of Eorzea worth giving up everything she used to be? She was cold, calculating. The ice to Danica’s flame. She couldn’t remember the last time she had truly smiled. She twisted her ring on her finger, absentminded, distracted. “I think he would have been proud of us for continuing on. For not letting his… his death stop us.” She hesitated, her breath catching in her throat for a moment. She looked back at the ring Danica wore around her neck, silently acknowledging it. Taking a deep breath, she poured herself another drink, downing it quickly. Haurchefant wouldn’t be proud of her for drinking her pain away. For trying to forget. Forget the look in his eyes as the life faded from them. She flinched as through she had been slapped. Halone help her.

Danica sat up, determined and inspired by her friends dour confirmation. Raising her glass, she began. “Well then, To us.” She started, extending arm in a toast. “May we keep making him proud.” The clink of glasses that followed rang hollow in their hearts, devoid of such an integral piece for so long.

#ffxiv #ff14 #haurchefant greystone #haurchefant x wol #danica voss #aveline de Bontensont #start #chapter 1 #Important

6 notes · View notes

daisychainblogs · 5 years ago

Text

Little Women - *snapshot review* & the taste of a team effort yet to come...

Greta Gerwig's films never fail to showcase visual attitude of formidably groovy personality: Frances Halladay skipped down a gloriously monochrome New York sidewalk, Lady Bird (Christine) McPhearson sauntered between the poplars of a golden and glittering Sacramento neighborhood - but Little Women ramps up the glory that is The Gerwig Aesthetic to a whole ... new ... level ...

First up: let's talk costume, designed by Jaqueline Durran - the mastermind behind the wardrobes of other period flicks Pride & Prejudice (2005), Atonement (2007), and 2020's triumphant 1917. Our ensemble of sisters ramble through their shifting landscape clad in a muddle of skirts and trousers; tam o’shanter hats and all engulfing scarves. Jo kicks about the hills and city boarding houses in handsome heavy-velvet waistcoats and berry-red capes; bloomers-for-trousers and stomping-leather-boots for shoe. Durran’s ability to show us character-in-cloth is a truly remarkable thing - we see Jo’s tomboyish inclinations in and through her stubborn, aptly askew jackets; we see Amy in her fur collars and extreme-circumference skirts, and the love-sick-Laurie in his mess of just-who-am-I-playing-today layers.

Jess Gonchor’s production design comes-up-roses on the set side of things also. The March house - chocka block with wooden floorboards, softly cushioned chairs and perfectly cluttered surfaces - is a bohemian jumble of complete wonder; of sibling squabbles and artistic pursuit, costume-stuffed-chests and rafter flung flower garlands.

And yet - despite all of its warmth and happy chatter and generally gorgeous guise - an elephant surveys Gerwig's adaptation from the corner of the room. Lifting a tentative trunk, Nelly quietly ponders - but why tell the tale of the March sisters again? ‘Aren't there new stories to be told?’, a friend quipped post viewing. Yes, she had liked it. No, she urged, there was nothing wrong with it. Still, she hesitated to fall flat at the feet of the Gerwig-Alcott colab. But for me, there has never been a better time for the March family to re-open their doors - and it is all about Alcott’s championing of choice.

The 21st century seems to worship almost exclusively at the cult of the extrovert, and to revel in the rungs of the career ladder far more than those of the ladder one might find in the garden of a family home. 'To be successful is to be dedicated to one's career motives'; 'To be likeable is to be cool, calm, and outgoing-ly collected', the modern world preaches.

It's not that Alcott doesn't praise the value of such attributes. She does, after all, give us the wonderful Jo - gloriously feisty, fiercely independent and reluctant to commit to a life of domesticity. But the genius of Alcott’s story - and the importance of its retelling - lies in the affection it shows for alternative choices.

"Just because my dreams are different from yours doesn't make them any less important", Meg informs her panicked-at-the-prospect-of-matrimony younger sister.

'To be successful is to be dedicated to one's career motives'; 'To be likeable is to be cool, calm, and outgoing- ly collected', the modern world preaches

Little Women reminds us that it's okay to be shy like Beth, or romantic like Meg. Amy’s social butterfly-ing is fantastic, as is Jo’s diametric opposition to 'elegant society'. It is fabulous to be quiet - and fabulous to be loud - and fabulous to be somewhere in the middle. The March sisters choose and pursue their own, individual definitions of joy. For Meg, it's a life alongside the man with whom she is in love; for Jo, joy is a piece of paper and an ink bottle perched on the desk of a New York city boarding house; Beth chooses the staves and an empty room, whilst Amy swirls her path through the ballrooms of high society. Marriage or career; socialite or social recluse: Gerwig trains a tender spotlight on Meg, Amy, Beth, and Jo,celebrating their own personal passions and desires with level enthusiasm.

In retelling Alcott’s story, Gerwig reminds us to celebrate choice; if men and women waking up to an 1800’s morning can show love to all the paths down which we might choose to travel - then so can we.

Gerwig's Little Women is a world of gleefully bickering sisters and gently reprimanding mothers; returning fathers and surrogate brothers; it is comfort and heartbreak, art and family, fighting and loving. It is the celebration of choice.

If only we could all stay there - tucked among the papers and paintings and pianos - for just a little while longer.

#little women #Greta Gerwig #costume #Film Review

3 notes · View notes

welovelofi · 5 years ago

Photo

Christmas is upon us and we brought you presents. Another round of Loves Me / Loves Me Not, with a little extra candy to last you through the holidays. Lets commence:

Neighbours Burning Neighburs - Grace

https://soundcloud.com/neighboursburningneighbours/grace

What better bandname to start off the holidays edition than this by Dutch punkrockers? It’s a perfect outfit to fill that void you’ve had in your soul ever since Sonic Youth was no more. This tune of theirs will instantly take you back to the late 80′s and early 90s with all the right teenage angst to cause a riot. This band is releasing their stuff on Dutch label Subroutine, who continue to bring forth class A acts, lately they’ve released songs by The Sweet Release of Death and The Homesick. Neighbours Burning Neighbours seems a wonderful fitting addition to their brilliant roster and I can’t wait to hear more by the band.

Modesta - Violet Fields

youtube

When you wake up after that long night and the hangovers are actually kind of nice and you’re feelin’ happy that you don’t need to have any plans for the rest of the day cause you’re too woozy to really do anything proper. This is most likely the first track on the album you should be putting on that day. It doesn’t require anything of you, but it’s gonna make you feel better about anything. If you’re going swimming, skiing or just taking a trip down the street on your electric kick-scooter - you know in a heartbeat this tune will make you look sexy no matter what you’re rocking <3

Postcards - Freediving

https://soundcloud.com/t3records/8-freediving/s-GOkJa

I really loved the Alvvays album Antisocialites that came out in 2017 - it’s a perfect testament to how you can make something as pleasant as worldly possible without giving up on aesthetics. This song by Lebanon artist Postcards is by far the song that comes the closest to capturing that same feeling, and I can only look forward to what other songs they might be releasing the next year time. They will be touring Germany in March, I would go if I was close.

Bauwaves - Early Morning Summer

https://soundcloud.com/bauwaves/early-morning-summer

This is the most gritty diy sound, I’ve heard in a long time, that is somehow also extremely pleasant to listen to - way too much distortion on everything in just the right way. Drums sounds terrible in just the right way. I’m not 100% sure how much of this was done on purpose - but the less it was the more authentic and chaotic it sounds. And this guys vocal is just fucking great! It could very well be a solo project - I’d love to find out? This is none the less as fresh as it comes. I’m keeping my eye on it for sure.

Will Wolfe - Saying Goodbye

https://soundcloud.com/manimal-vinyl/will-wolfe-saying-goodbye/s-EkYR0

Lets pretend that the 80s glam was all made up of people wearing cowboy outfits. All the same sass but everything in a “Western” bubble. Id definitely fucking go to the champagne parties being awkwardly fancy. My biggest wish would probably be to run into Will Wolfe so I could ask him how he created this world of his and how we can all learn to mix things so wonderfully for the future? until this happens I hope there’ll be some videos out by this guy and his brilliany 80′s Sassy-Western <3

Maxime. - Arm & Hammer

https://soundcloud.com/maximemakesmusic/arm-hammer

Pretty damn nice lofi bedroom tune complete with full yet shoddy midi-drums and early Bon Iver vocals. On top of this it’s just a pretty nice tune in itself which is why this little lazy monday gem should be on your playlist for when you’re feeling off. I can recommend it for laying down pretty much anywhere, on the floor at the trainstation with people rushing by or in your bathtub full of bubbles. Eyes closed or not it’s a winner.

Bled Tape - Black Ice

https://open.spotify.com/track/25VEyGNcJ3WuixjDjswdXR

Second time we have Bled Tape on here and for good reason. There’s something very very right about the lofi aesthetics, vocals drowning in everything, instruments sounding like it was just fresh out of the bedroom. The only thing I’d love to hear is the drums less midi perhaps - it has a nice vibe to it, but I’d think with just a bit more effort this could really go down a wonderful path. It’s not to keep you from wasting any time digging into this act, there’s good indication that there’s more great tunes to come. It’s like all the 90s angst are hanging onto this tune weighing it down to an almost slowmotion pace! Well done, certainly.

baklavaa - dsnylnd

youtube

Probably one of the coolest music videos I’ve seen in a long time, the only one rivalling it would be Potty Mouth by the Malpractice> https://www.youtube.com/watch?v=KWsJJLqndOg - a video in colab with artist Joe Pascale - I recommend you to check it out, but only after you gave this video and tune a look. To me this is nothing shy of Butthole Surfers anno 2020, the music has all the insane diy-like elements of BS and the video is the chaos of our time crammed into a 4min youtube display. I’m pretty sure Gibby Haines and the rest of the band would be proud. Just sit with this tune and video a while and realize how nice it is. Some might cast a first glance and dismiss it as too off - you’re missing out, your loss not mine.

Mother Sun - Happily Sinking Heart

https://soundcloud.com/mothersunmusic/05-happily-sinking-heart

Back to the feel good tunes of 2019, Mother Sun created a laid back tune that hints of country but without all the annoying parts of it. Like in Perfume by Patrick Süskind you get an insight into how they distilled scents, this song probably has 20 albums of bad country songs it was distilled from- It’s not that I’m hating on country it’s just there’s so much silly nonsense there with too many people that have nothing to say. I don’t really care that you went to the restaurant with Dorothy this afternoon and came home to hang out on your porch, just please have more to tell me than that Peter. But I’d venture into the world of the American south with this track on my ghettoblaster. Perhaps I’d even make a friend or two walking down the street pointing a shooting finger or two at passer-bys with a wink and a smile, I’d prefer it to happen in slowmotion, but you can’t have it all of course. Check out the tune for your monday mornings, you won’t regret it.

Nathan & Jessie - Mi Corazon

youtube

We all know it’s not Christmas without a little love, so from us to you here’s your 2019 Christmas tune. It’s guaranteed to keep you warm even through the coldest of winters. If you’re in a remote ski-lodge somewhere in Canada with 5 feet of snow to dig yourself out of, just pop this on and you’ll be south of the American border in no time. You can thank us later, or just thank Nathan & Jessie for this wonderful Christmas classic that should be on all the damn streaming hitlists any day of december. Mariah Carey had a good run, it’s time to pass the torch Mari!

#Neighbours Burning Neighbours #Modesta #Postcards #Bauwaves #Will Wolfe #Bled Tape #Nathan & Jessie #Mother Sun #baklavaa

3 notes · View notes