tasty-patches

[tasty-patches]

Ah, I see. I couldn't leave a reply because I had ur-daily-inspiration blocked. Probably for reposting uncredited photos.

Fairy flowers appreciation post

[tasty-patches]

Tumblr decided to post this instead of saving a draft while I was halfway through typing, so if you saw an incomplete version, the version I'm reblogging is the complete one.

Also my partner says she could leave a reply, but the reply button is grayed out for me and it says replies are restricted, so I'm not sure what's up with that. More Tumblr weirdness. It ended up way too long to put into a reply anyway, so whatever, it worked out.

Fairy flowers appreciation post

[tasty-patches]

Well I would have preferred to make this comment in the replies, but replies are turned off, so...

These are (almost certainly) not AI. They don't exhibit any hallucinations, inconsistent patterns, or impossible depth of field. These are also all very clearly real flowers.

If you don't want to read any more of this post, then take it from a photographer who has taken a lot of photos of flowers. I do not believe that these are AI. Current AI models (in November of 2024) would probably not be able to produce images of this quality with petals as consistent as these. At least two of these are 2+ years old and models from two years ago produced images that looked like plastic. Read on for what each image depicts and some cursory info gleaned from quick reverse image searches.

The first image is a hibiscus. Most likely the hibiscus x rosa-sinensis hybrid (Chinese hibiscus, Hawaiian hibiscus, shoeblack, among other names) and as far as I can tell is pulled from an Amazon listing for seeds of the cultivar.

The second image is of nelumbo nucifera (sacred lotus, Indian lotus), of which there are a handful of high petal-count varieties. The oldest version of this photo I can find is on a Vietnamese health and lifestyle blog from August 2022. If you don't remember what AI generated flowers looked like two years ago, they looked like Fisher-Price toys. I doubt a model today could make that coherent of an image, but they certainly couldn't in 2022.

The third image is a lily of some variety. Probably a Stargazer lily which have a white fringe on the petals that responds quite well to stem dyeing. I can't find an original source for the photo, but it was being passed around Pinterest as early as July 2022. Again, an AI image generator at the time would not have been capable of producing an image of this quality.

The final image is also some sort of lily, specifically a double-flower or double-bloom cultivar of an Oriental lily or one of its hybrids. Like the others, I can't find an original credited source, but the image appears to have been pulled from an Amazon listing for lily bulbs.

Now, after establishing that these are probably not AI, there are two things I want to discuss. Why people might think that legitimate photos (including product photos in this case) are AI even though they aren't, and why crediting sources for photos is not only ethically correct but can avoid this kind of confusion and ensure people don't mistake legitimate art for AI.

(I love Tumblr, I don't hate it, even when it posts a draft I was still working on and deletes three paragraphs I had already written. I love Tumblr, I don't hate it....)

I'm colorblind (blues and purples look the same; orange, red and yellow all appear green or brown depending on the saturation and value; teal, cyan and pink all appear gray or white depending on value) so I can't comment on the color of these specific photos, but I can tell you how I approach color in my own.

Photos are art. When I process a photo of a flower, I'm not trying to produce an image for a botany textbook. I'm trying to get the viewer to perceive the composition with the same feelings and attention that I had when I took the photo. I will exaggerate saturation, vibrancy and contrast if it makes a flower look more like a flower. I will desaturate greens to be almost gray to make sure leaves, stems and grass don't distract from the subject.

These adjustments produce an "unnatural" image, but again, the photo is art. It's meant to convey a feeling and a one-to-one, life-accurate portrayal of a subject doesn't convey what I felt when I looked at it, it just feels soulless.

If people look at a photo of a flower I took and they don't believe that it's a real photo, then maybe that's a failing on my part as an artist, but at least I have a raw photo and all of the metadata to prove that I took that photo with one of my cameras sitting on my desk right now.

Which brings me to my last point. I can't find verifiable original sources for any of these photos. I don't know if the Amazon listings are original uploads from the store owners or if they just found images on Google search and used those, or if they were stock photos taken from Adobe, Getty or any of the other stock photo services.

I'm not here to shame anyone for sharing unsourced photos, especially when sources aren't readily apparent, but it's sad to see it happen and it would have been better if sourced images of photos had been used instead of four nice looking images from a Google search of "pretty flower picture."

If legitimate photos can't be traced back to a source, it will become harder and harder to tell the difference between AI and photography. For some types of images, especially ones with a lot of non-repeating information, AI is already producing images of nearly identical quality to heavily postprocessed photos. They still struggle with repetitive patterns and text, but they're getting better and it's only a matter of time.

The only way to make sure that photos can be verified as real is to maintain a credit line for every photo. If you take photos and repost then somewhere without credit, not only are denying an artist attention for their work (if not outright claiming it as your own), you're denying viewers an opportunity to verify the authenticity of the work.

I would prefer that people not repost photos at all and instead link to original sources, but I've been around on the Internet for a long time and I know that that's not a realistic expectation. Maybe photographers should all be watermarking their photos, but I personally hate the look of watermarks and signatures on my own work. I don't even sign the front of my paintings.

Anyways, thanks for reading my ramble, have a nice day.

Fairy flowers appreciation post

[tasty-patches]

Me knitting: Haha fuck yeah!!! Yes!!

Me purling: Well this fucking sucks. What the fuck.

[tasty-patches]

Well, things may be Not Tremendous right now, but at least we got good enough weather to nail down an autumnal-ish Melina photoshoot? I've still got a few things I'd like to touch up, but I do love her~ And what a comfy costume to wear while tromping around the woods!

Photos by @tasty-patches

[tasty-patches]

Yeah. Me too.

[tasty-patches]

Excalibur (1981) dir.: John Boorman

[tasty-patches]

You don't need a "WIFE". What you NEED is a beautiful priestess of the old religion to lead you down a dark path.

[tasty-patches]

[tasty-patches]

Just a few more hours

[tasty-patches]

Cocaine is also people

[tasty-patches]

[tasty-patches]

Hi, I set up a Printify store! It's a little sparse right now, but I'll be adding more designs soon.

One very limited item is my Figayda fanart on a sherpa blanket. I wanted one for myself and didn't want sapphic enthusiasts to be jelly, so it's listed until November 10th.

[tasty-patches]

Melina is (almost totally) DONE, WHEW, so I wanted to get some preview-y detail photos in celebration.

(Photoshoot will HOPEFULLY be soon...and there will probably be nicer detail photos to go with that, but whatever. :P)

[tasty-patches]

Inktober Day 23: Domino

I wanted to do a take on a design I really hated, lol.

[tasty-patches]

Don't go shower without her

[tasty-patches]

Inktober Day 13: Zatanna ✨