#ai transcribe audio

[ephemerasnape]

Bred by Victor Rookwood (Audio)

This is what it's all been leading up to.

🎩Rookwood x You

Admit it, little witch.. You're dying to be impregnated by this man. And now daddy is going to give you what you so desperately desire. F!Listener

"When daddy comes in you, he means it."

EXPLICIT AUDIO 18+

Breeding / Impregnation / Dominance / Daddy Kink / Praise Kink / Pregnancy Kink / Age Difference / Dirty Talk / Smut / Excessive Talking During Sex / Bad Latin

Not sure whose dirty talk is hotter - DR or @rookwoodswife!

Transcript beneath the cut

Victor Rookwood:You’re mine now darling. Now that I’ve gotten my hands on you I'm never letting you go. No. You belong to me in every way. Physically, mentally and emotionally. 

Kissing, muffled moans

Victor Rookwood:Every part of you is mine. Including your womb. 

Muffled moans and kissing

Victor Rookwood:Yes darling, you’re going to have a lot of children for me. Beginning immediately. 

Listener: Fuck

Victor Rookwood: Oh yes, I’m going to fuck you every night from here on out princess. And I’m going to make sure it takes.

Listener: Oh gods, yes.

Victor Rookwood: Did you think I was fooling around? Victor Rookwood doesn’t play games, little witch. When daddy comes in you, he means it. So just lay back and spread your legs for me

Buckle coming undone, moaning, skin slapping against skin rhythmically

Victor Rookwood:Your pussy is so tight. Daddy’s going to fix that for you.

Victor Rookwood:That’s it. Take my cock. 

Victor Rookwood: Let daddy take his pleasure from that tight little cunt. Fucking you hard and deep. Pumping his cum into you over and over again.

Listener: Fuck 

Victor Rookwood: Filling you. Breeding you. Admit it darling, you want to be bred by me.

Listener: Yes, daddy.

Victor Rookwood: You want to have my baby.

Listener: Breed me daddy

Victor Rookwood:Oh gods... Beg for my seed darling. 

Listener: Please, I need your cum.

Victor Rookwood: Good girl. Daddy’s going to fill you with so much cum. Tell me how badly you need it

Listener: I need it so bad, please

Victor Rookwood:Tell me you want to have my baby.

Listener: Please, I want to have your baby. 

Victor Rookwood: Daddy’s little girl is so obedient.

I can’t wait to watch you swell, heavy with my child. You’re going to be so helpless, so vulnerable. Marked as mine. For everyone to see.

Listener: Oh gods, yes

Victor Rookwood: With just a single glance at you, everyone will be able to tell that I fucked you. You know that right?

Listener: Yes, daddy.

Victor Rookwood: You’re about to become so thoroughly mine, in every way. Are you ready, little witch?

Listener: Yes Daddy, come in me

Victor Rookwood:Good, because it’s coming. Fuck yes. Groans. Take all of it. Every. Last. drop. 

Listener: Gods yes. I love you.

Victor Rookwood: Good girl. Now, we’re not leaving anything to chance.

Radico!

Well darling, what do you say when daddy impregnates you?

Thanks so much to @berserkerrose for the transcription!

[carigm]

The “Teaser”, mlvn rooftop convo, and Lord of the Rings parallel…

This is gonna be a long post, so grab some snacks y’all.

First of all, the teaser the Duffers shared at this Netflix shareholders event was basically all BTS stuff, and according to someone that was there, we have mostly seen all of it. The actual clips from the show they showed were so short that most people missed it. However, over those short clips it seems they played a voiceover of part of the mlvn rooftop convo. Notice how the Suffer Sisters are literally incapable of sharing anything new, and the only audio they disclosed is from the ONE scene that’s been leaked to death, and even transcribed multiple times with the help of AI. In any case, Netflix did not share this teaser with the masses, and it’s unlikely they ever will. Stranger Things is not going to the Super Bowl this year (yes you heard that right) and the Tudum Event isn’t until May. Our only hope before that would be them releasing something on Will’s birthday, but whether in March or May, I believe we’ll be getting a proper teaser by then.

People that attended the event reported that El has a voiceover line where she goes “they don’t get to write the ending, we do” and apparently a voiceover Mike line where he goes “we’ll finish this together” (I’m not sure if this was paraphrased or not). Immediately, we all realized that these lines sound pretty close to what Mike is allegedly saying to her during the rooftop scene. Many people in the fandom have taken the time to transcribe that scene, some with AI and some without, and although some things could be wrong here and there, the general idea of it seems pretty clear. I’m attaching an AI reading of the scene here, so I can point out where I think his dialogue might be from…

Around the 1.43 mark, the AI picked up “enjoying it, together”, however I think this is where the “we’ll finish it, together” line comes into play. If anything, this shows AI isn’t 100% accurate, and it does call into question some of these previous lines 😂…I think it’s possible that after this speech from Mike about stories, fantasy endings and heroes, El tries to follow his advice and be positive, and maybe she delivers the “they don’t get to write our ending, we do” line back to him. It’s unfortunate because obviously we can’t see her face in the video, but I think it makes sense she would reply with that because right after it seems like he says “of, course…” and then proceeds to seemingly add that the Party can have a happy ending, without all the fantasy elements he mentioned before.

I find it very interesting that he’s choosing to speak to her with this storytelling analogy, which at first I believed to be a D&D analogy, but the more I think about it, the more I feel like he’s talking about an actual story. And then the lord of the rings parallel hit me, specifically with this scene. If you’re not aware, Finn Wolfhard has mentioned lotr twice now when talking about season 5, and I personally think it’s possible that Mike is using lord of the rings here as a reference to describe the hero’s journey and relate it to what the party has been through. Think about it, he’s trying to cheer El up, who has been stuck in that fuckass radio station for a year, who’s probably extremely tired of everything she has been dealing with for years, and he just wants to offer her some consolation so she can keep going and fighting. Does that sound familiar?

Well my friends, if it does, that’s because it is a direct parallel to Frodo and Sam from Lord of the Rings. I’ve always thought Byler were insanely samfrodo coded (funny enough the last S4 Byler scene is almost identical to this scene too), but it seems the Duffers are paralleling mlvn to them here. In lotr, Frodo bears the biggest burden of the story, as he follows his hero’s journey to Mordor to defeat evil. Along the way, ofc, he becomes increasingly weary and hopeless, and it is up to Sam (his best friend) to cheer him up and provide him with strength to keep him going. How does Sam do this? Interestingly enough, he encourages Frodo by describing all the beautiful things that will come AFTER they have won, what they and their friends will be able to enjoy when they get back home. Basically everything Mike appears to be saying to El in this scene, fantasizing about the end of the battle. To make the parallels even crazier, while on his hero’s journey, Frodo has to remain in hiding because there are multiple forces looking for him, and we know that El is basically hiding away from the government.

Another thing I want to point out is that in lotr (spoilers I guess 😭) good does win in the end, and the main characters get to return back home. However, Frodo is so changed by the journey and all the things he encountered that he simply cannot stay with his friends. Instead, he leaves and goes to the Undying Lands, where he finds peace. He doesn’t die, but he also cannot stay in Middle Earth. Him and Sam have a beautiful goodbye scene and then Sam is left with the literal book of stories Frodo started, and is told by Frodo to “finish it”.

Make of that what you will…

[mariacallous]

On Saturday, an Associated Press investigation revealed that OpenAI's Whisper transcription tool creates fabricated text in medical and business settings despite warnings against such use. The AP interviewed more than 12 software engineers, developers, and researchers who found the model regularly invents text that speakers never said, a phenomenon often called a “confabulation” or “hallucination” in the AI field.

Upon its release in 2022, OpenAI claimed that Whisper approached “human level robustness” in audio transcription accuracy. However, a University of Michigan researcher told the AP that Whisper created false text in 80 percent of public meeting transcripts examined. Another developer, unnamed in the AP report, claimed to have found invented content in almost all of his 26,000 test transcriptions.

The fabrications pose particular risks in health care settings. Despite OpenAI’s warnings against using Whisper for “high-risk domains,” over 30,000 medical workers now use Whisper-based tools to transcribe patient visits, according to the AP report. The Mankato Clinic in Minnesota and Children’s Hospital Los Angeles are among 40 health systems using a Whisper-powered AI copilot service from medical tech company Nabla that is fine-tuned on medical terminology.

Nabla acknowledges that Whisper can confabulate, but it also reportedly erases original audio recordings “for data safety reasons.” This could cause additional issues, since doctors cannot verify accuracy against the source material. And deaf patients may be highly impacted by mistaken transcripts since they would have no way to know if medical transcript audio is accurate or not.

The potential problems with Whisper extend beyond health care. Researchers from Cornell University and the University of Virginia studied thousands of audio samples and found Whisper adding nonexistent violent content and racial commentary to neutral speech. They found that 1 percent of samples included “entire hallucinated phrases or sentences which did not exist in any form in the underlying audio” and that 38 percent of those included “explicit harms such as perpetuating violence, making up inaccurate associations, or implying false authority.”

In one case from the study cited by AP, when a speaker described “two other girls and one lady,” Whisper added fictional text specifying that they “were Black.” In another, the audio said, “He, the boy, was going to, I’m not sure exactly, take the umbrella.” Whisper transcribed it to, “He took a big piece of a cross, a teeny, small piece … I’m sure he didn’t have a terror knife so he killed a number of people.”

An OpenAI spokesperson told the AP that the company appreciates the researchers’ findings and that it actively studies how to reduce fabrications and incorporates feedback in updates to the model.

Why Whisper Confabulates

The key to Whisper’s unsuitability in high-risk domains comes from its propensity to sometimes confabulate, or plausibly make up, inaccurate outputs. The AP report says, "Researchers aren’t certain why Whisper and similar tools hallucinate," but that isn't true. We know exactly why Transformer-based AI models like Whisper behave this way.

Whisper is based on technology that is designed to predict the next most likely token (chunk of data) that should appear after a sequence of tokens provided by a user. In the case of ChatGPT, the input tokens come in the form of a text prompt. In the case of Whisper, the input is tokenized audio data.

The transcription output from Whisper is a prediction of what is most likely, not what is most accurate. Accuracy in Transformer-based outputs is typically proportional to the presence of relevant accurate data in the training dataset, but it is never guaranteed. If there is ever a case where there isn't enough contextual information in its neural network for Whisper to make an accurate prediction about how to transcribe a particular segment of audio, the model will fall back on what it “knows” about the relationships between sounds and words it has learned from its training data.

According to OpenAI in 2022, Whisper learned those statistical relationships from “680,000 hours of multilingual and multitask supervised data collected from the web.” But we now know a little more about the source. Given Whisper's well-known tendency to produce certain outputs like "thank you for watching," "like and subscribe," or "drop a comment in the section below" when provided silent or garbled inputs, it's likely that OpenAI trained Whisper on thousands of hours of captioned audio scraped from YouTube videos. (The researchers needed audio paired with existing captions to train the model.)

There's also a phenomenon called “overfitting” in AI models where information (in this case, text found in audio transcriptions) encountered more frequently in the training data is more likely to be reproduced in an output. In cases where Whisper encounters poor-quality audio in medical notes, the AI model will produce what its neural network predicts is the most likely output, even if it is incorrect. And the most likely output for any given YouTube video, since so many people say it, is “thanks for watching.”

In other cases, Whisper seems to draw on the context of the conversation to fill in what should come next, which can lead to problems because its training data could include racist commentary or inaccurate medical information. For example, if many examples of training data featured speakers saying the phrase “crimes by Black criminals,” when Whisper encounters a “crimes by [garbled audio] criminals” audio sample, it will be more likely to fill in the transcription with “Black."

In the original Whisper model card, OpenAI researchers wrote about this very phenomenon: "Because the models are trained in a weakly supervised manner using large-scale noisy data, the predictions may include texts that are not actually spoken in the audio input (i.e. hallucination). We hypothesize that this happens because, given their general knowledge of language, the models combine trying to predict the next word in audio with trying to transcribe the audio itself."

So in that sense, Whisper "knows" something about the content of what is being said and keeps track of the context of the conversation, which can lead to issues like the one where Whisper identified two women as being Black even though that information was not contained in the original audio. Theoretically, this erroneous scenario could be reduced by using a second AI model trained to pick out areas of confusing audio where the Whisper model is likely to confabulate and flag the transcript in that location, so a human could manually check those instances for accuracy later.

Clearly, OpenAI's advice not to use Whisper in high-risk domains, such as critical medical records, was a good one. But health care companies are constantly driven by a need to decrease costs by using seemingly "good enough" AI tools—as we've seen with Epic Systems using GPT-4 for medical records and UnitedHealth using a flawed AI model for insurance decisions. It's entirely possible that people are already suffering negative outcomes due to AI mistakes, and fixing them will likely involve some sort of regulation and certification of AI tools used in the medical field.

[bitchesgetriches]

Why we’re against AI as a writing tool

Sophisticated AI tools like ChatGPT are the result of systemic, shameless theft of intellectual property and creative labor on a massive scale. These companies have mined the data of human genius… without permission. They have no intention of acknowledging their stolen sources, let alone paying the creators.

The tech industry’s defense is “Well, we stole so much from so many that it kinda doesn’t count, wouldn’t ya say?” Which is an argument that makes me feel like the mayor of Crazytown. I don’t doubt the courts will rule in their favor, not because it’s right, but because the opportunities for wealth generation are too succulent to let a lil’ thang like fairness win.

I’m not a luddite. I recognize that AI feels like magic to people who aren’t strong writers. I’d feel differently if the technology was achieved without the theft of my work. Couldn’t these tools have been made using legally obtained materials? Ah, but then they wouldn’t have been first to market! Think of the shareholders!

We’re lucky to have the ability and will to write. We won’t willingly use tools that devalue that skill. At most, I could see us using AI to assist with specific, narrow tasks like transcribing interview audio into text.

At a recent industry meetup, I listened as two personal finance gurus gushed about how easy AI made their lives. “All my newsletters and blogs are AI now! I add my own touches here and there—but it does 95% of the work!” Must be nice, I whispered to the empty void where my faith in mankind once dwelt, fingernails digging into my palms. It’s tough knowing I’m one of the myriad voices “streamlining their production.”

I feel strongly that every content creator who uses AI has a minimum duty to acknowledge it. Few will. It sucks. I’m frothing. Let’s move on.

Read more.

[remyfire]

Much as it pains me to do, I must ask once more for a little bit of help to pay for Peach's medical care for this month.

Peach is my wonderful, beautiful, perfect, sassy, cuddly, loving cat. She's 12 years old, and a few months ago we found out that she is diabetic. I love her so, so much and I will do anything I can to keep her healthy and give her a wonderful life as she enters her senior years, but unfortunately the costs of caring for a diabetic cat are more than I can afford right now.

Thanks to the AI boom, the transcription industry is drying up because if there's one thing these things are good at, it's transcribing audio files. Work has been increasingly difficult to come by—I have yet to receive any for this week—and while I've been fortunate enough to have a handful of job interviews recently, I might not hear back about if I got a position until the end of September. And wouldn't you know it, but my credit card is just a hair away from maxing out.

Here's the current damage, which includes Peach's prescription food ($67), syringes ($19), and a fresh bottle of insulin ($133). Cats can have their diabetes go into remission, and if that happened then we would only need to maintain her diet, but in order to make that a possibility, I need to be meticulous about her care.

I've set up a Kofi goal here ($230 in order to cover PayPal fees) so you can see how much I still need for her cost of care. I know that this is a miserable time to ask for donations and that we are so fatigued on giving already, but I legitimately have no choice. Until I can find work somewhere, I desperately need help.

Reblogs are appreciated. And here's a video of her playing with a shoelace as thanks for taking the time to read this post!!

[disbabeled]

To everyone making New Years Resolutions right now:

Please consider making it a resolution to create accessible content.

[Plain Text: Please consider making it a resolution to create accessible content.]

When you make art, consider making an image description and/or alt text to go with it.

When you share videos, consider adding captions or subtitles.

When you post memes or just images in general, consider adding image descriptions/alt text to them.

There is so much content out there but so little of it is actually accessible. It doesn't take much time at all to make it accessible, either. Especially since there's already resources and softwares out there to get you started.

This link discusses the differences between alt text and image descriptions as well as how to create them on Tumblr (Note: Some info is outdated. Alt text can now be added/edited after the image is posted).

If you don't want to make captions yourself, I've heard good things about Hitpaw Edimakor. You can also use apps such as LiveTranscibe on Android and even Google Docs has a function for turning audio into text.

Even if you won't make alt text or captions or whatnot yourself, just taking a look through the reblogs of a post to find an accessible version is helpful.

Just make sure to double check any content before sharing it as many people will add alt text to images as a joke, meme, or just as a way to include more information about the image (Rather than as actual alt text). This goes double if you're using AI or other softwares to transcribe/describe/otherwise make content accessible.

Now go forth and make shit accessible!

[crimeronan]

kitkat's dirt-cheap writing and editing services!

hello all! i'm kitkat, i'm a professional writer and editor who's been professionally writing and editing for well over ten years. as far as tumblr audiences go, though, i'm better known for my work over at ao3 elliptical.

the vast majority of my prior clients have dropped me for AI, and i need to build my freelancing base up again. so i've created a fiverr profile.

however, in order to get work on fiverr, you need good reviews. and before you have good reviews, you need to offer your services cheaply enough for people to take a chance on you.

so! i have some dirt-cheap offerings at the moment (11/14/23):

writing 500 words for $5

writing 1,100 words for $10

writing 1,700 words for $15

editing 1,000 words for $5, $10, or $15 (depending on how in-depth you go)

transcribing audio/video (price starts at $5 and depends on length)

my MAIN ask if you take advantage of the offer is to give me a five-star rating and nice review. if just a couple people here buy my stuff, i can show fiverr that i actually am good at my job. and then they'll show my profile to more people, and i can work up to a living wage!

quick faqs:

can i buy fanfic from you?

i cannot legally write fanfic for money. however, if there's a certain AU, relationship, character, prompt, etc., that you want to see, you can tell me (there should be a free-write space to explain what you're looking for when you order). then i can write you something that's "like" that, just with original names.

can i pay for you to edit/beta my fanfic or original fiction?

ABSOLUTELY, yes. just keep in mind that only the $15 tier currently includes in-depth comments and constructive feedback. with the $5 and $10 tiers, i'll happily proofread and check your grammar, but i won't have a ton of actionable feedback to give!

what's the copyright situation on this stuff?

you have the full rights to anything you buy from me on fiverr, forever. more in-depth explanation here!

i can't buy anything right now but want to support you. how can i do that?

it is 100% okay if you don't have money to spend (or simply don't wanna buy my writing)! one helpful thing you can do is favorite my profile, if you have a fiverr account. you can also make a fiverr account pretty fast if you have a gmail, but it is completely reasonable not to want to do that.

you can also reblog this post if you want! i think most of my potential audience for this already follows me..... but if anyone wants to sing my praises in reblogs, i definitely won't say no, LOL

[pinene]

What they should do is let AI pick out individual sentences or words of speech in a video and then feed those to a human contributor who transcribes them. this would make captioning really easy since the timing isn't something you have to nail, you just have to transcribe given snippets of video (though likely in the context of the larger video as that's important to transcription)

is this a thing? I hope it is. I'm imagining that it essentially plays you the video and then is like "okay what did that person just say?"

And I also think this would be a good example of human-AI synergy. Because it's much easier for machines to learn how to pick out human speech in an audio data stream, because data is the context they live in. But AI being able to exactly pick out the words someone's saying in a video? Completely out of its context. And it's tougher for humans to be pausing and playing and scrubbing the video to time captions, because we have to interact with data thru intermediary layers. But we can easily listen to the CONTEXT and TONE of the video and use it to pretty accurately transcribe what's being said. that should be our job !

[walks-the-ages]

Sorry peeps, but if you're genuinely out here trying to defend generative Ai because you think anyone against it is "ableist" sorry not sorry you're not just getting unfollowed you're getting fully blocked along with the OP and who you reblogged it from lmao.

"most people angry about Nanowrimo allowing AI are just being loudly ableist!!! Generative Ai is a great tool for disabled people!! Everyone should be able to use it!"

Hmm, sorry, maybe you need to curate your dash some more like I just did by blocking you, but literally the ONLY people I've seen talking about Nanowrimo's AI stance are people who are actually disabled themselves who are pointing out how fucking shitty it is for Nanowrimo to defend themselves and their sponsor using AI (and possibly scraping your works to further train their AI) By using ~Disabled People~ in concept as a shield against criticism.

Many, many people have posts on here about how they are physically or mentally disabled and they would absolutely hate having someone belittle them by telling them the only way they can accomplish something creative like writing a novel is to have a Computer spit nonsense out into a word document, or generate a "masterpiece" digital image for them from a few words typed in...

Like.

If you actually care about disabled people, you wouldn't be advocating for generative AI to be used to erase their creativity by just letting a computer churn out crap.

If you can type in a prompt on an AI generator, you can type in a word processor to write your story.

Can't physically type at all?

Use speech to text,

or do an audio recording of your novel, and have someone transcribe it,

or use actual existing Closed Caption technology to transcribe it for you!

These are all accessible technology options that actually help disabled people be creative, not just tell an AI generator "hey write me a book about x"

Disabled people have been authors and artists for millennia.

Stephen Hawking used a combination of Predictive Text, eye-control cursors, and an infrared sensor mounted on his glasses that would detect if he was tensing or relaxing the muscles in his cheek, allowing him to scroll a virtual keyboard.

Somehow, I don't think the people championing generative AI actually care about "disabled people" when they try to insist that typing a prompt into a generator and having it churn out random slop is the solution to 'allowing disabled people to be creative' instead of actually giving them the various technology and accessibility tools that have been a thing for at least 25 years, like:

Eye-tracking software that allows you to type or paint on a computer screen (this is now at the point where people can play online video games with this software!)

Having any kind of smart phone set up with speech to text and a word processing app like Google Docs or a notepad app

Using basic sound recording apps to dictate your novel for later transcription

Using other body parts than your hands (or using prosthetics) to hold paint brushes, pens, markers, digital stylus, computer mouse, etc to make art with.

And so much more!

The real ableism here is when pro-AI bros try to insist that Disabled People, categorically, are incapable of being creative and accomplishing anything without a computer doing all the work for them by generating things based on millions of stolen works, and the complete erasure of all of the disabled artists who are here *now* and existed in the past, acting like they do not and never existed, all so that rich white ai bros can continue to flood reddit with "super cool badass art I just made" which is a nonsense amalgamation, and throw tantrums when artists start using programs like Glaze and Nightshade in an attempt to protect their art from those same predatory ai tech bros.

Technology is meant to help humans be creative, not steal our works and livelyhood by replacing writers and artists entirely, because all some rich guy has to do now is type in a prompt.

[martintoms]

It's my birthday! Born 8th January 1999. Today I'm turning 26, time flies, but my love for Fallout remains. I love this song so much that I learned to work with composing software and painstakingly spent hours putting the notes in! To my knowledge this is a unique, but completely lore-friendly piece. Composed by the Nuka Cola Corporation, transcribed by me, and played by audio magic of MuseScore :) This piece will be included in my mods at some point, Modus Operandi, looking at you :) No AI was used in the making of this piece. Just my human hands, ears and MuseScore. #nukacola #nukaworld #fallout4 #morgosus #martintoms #birthday #tutorial #sheetmusic #beginners #beginner #piano #music #jingle The Nuka-World Song (Beginner-Friendly with Piano Sheet Music) - 1st Birthday Special published first on https://www.youtube.com/@MartinToms/

[ephemerasnape]

Just A Little Blood, Redux (Audio) 🔞🔞🔞

Finally redid this audio at long last. I replaced a lot of horrid "MC/You" sound effects with better ones from @rookwoodswife. I also made it slightly less nice. It's still not my best work but it's unique in its own way.

Rookwood gets ahold of you and punishes you -- with his wand and with his cock. RAPE "This is for Selwyn. Morgan. Travers. All the others you slaughtered without a thought."

Victor Rookwood x You/MC

DARK & DISTURBING THEMES

EXPLICIT AUDIO 18+

F!Listener / Rape / Torture / Daddy Kink / First Time / Loss of Virginity / Age Difference / Dirty Talk / Villain Diatribe / Dead Dove Do Not Eat / No use of Y/N or MC but you are MC

Transcript beneath the cut!

Rookwood: Crucio!

You/MC:  Cries in agony

Rookwood: Where’s your ancient magic now?

Do you realize how many of my men you’ve killed, not to mention the goblin helm and dragon you stole? Or do you just kill so casually you've already lost count?

You/MC: Please, I’m sorry!

Rookwood: It’s far too late for apologies, little one. You’re gonna take whatever daddy wants to give you

You/MC: Whimpering,

the sound of a buckle being undone

You/MC: Cries in pain/surprise

sounds of skin slapping together rhythmically. 

Rookwood: Aw sweetheart, I know it hurts, but you’re making daddy feel so good.

Just a little blood. Nothing a strong obstinate little girl like you couldn’t handle. 

Groans Take it.

This is for Selwyn. Morgan. Travers. All the others you slaughtered without a thought.

For Rookwood Castle, my family home which you violated.

For Horntail Hall, all the lives lost there. All the poacher children you orphaned. This is nothing more than you deserve. 

Fuck! Gods yes.

*Speed of slapping increases*

*Rookwood groans. Skin slapping together slows and stops.*

Mmm.. *heavy breathing*

*He lights a match, inhales* 

Now that that unpleasantness is out of the way...

*exhales*

What am I to do with you?

---

Transcription courtesy of @berserkerrose - thank you!

[tofueggnoodles]

My Drama CD translation workflow

The first Drama CD I ever translated is One-Day Store Manager. That was in July 2022, nearly two years ago. Until recently, the workflow was simple, at least on the surface. I listened and either typed out the first thing that came to mind or looked up what I heard in jisho. If the sentence was long and complicated, I would listen to it at half-speed a few times, transcribed it into romaji and tried to figure out the translation as I stared at the transcription. Obviously, it was a slow and painstaking process.

During the past few years, rapid progress has been made in the fields of machine transcription and translation. Even back in 2022, I had the nagging feeling that things could be sped up.

Last April, I finally did it. After scouring the world wide web, I cobbled up some python scripts to do a couple of things.

1. Automatic transcriptions from mp3 files. Yep. This removes the major pain in the neck. However, as they say in computer science, garbage in garbage out (GIGO). The quality of the transcription depends on the audio quality of the mp3 file, of course. The amount of corrections I have to make is correlated to the bitrate. If the quality is good, I only have to correct about five percent of the transcription.

2. Automatic translation. Not so great, but it takes away the hassle of translating simple sentences and stock expressions. It is equivalent to copying and pasting into Google Translate site, but in bulk. Again, the GIGO principle applies. Hence the importance of making sure the Japanese text makes sense in the first place.

I will use a short scene from Kyouka Suigetsu to demonstrate the advantages and the pitfalls of automation in Japanese to English translation.

The speech to text output is a chunk of text, sans punctuation and indication of who is speaking. The script is particularly bad at transcribing our guy’s names. Hakkai is either eight floor, destruction, eight times or Bajie. (The last is not technically wrong, though.)

Here is how the translation looks like. Of course, it is also a chunk of text.

Here is the edited transcription. I corrected the obvious mistakes and indicated who is speaking at the beginning of their lines.

Again, GIGO. The output of machine translation this time is much better. It is not perfect, but I find it so much easier to re-translate with the Japanese text just below each line.

Still, the fact that even the automatic translation from the automatic transcription makes more sense than a translation made by a human being nearly twenty years ago is a testament to how far the transcription and translation technologies have come. The fears expressed by some professional translators are far from unfounded.

The day Google Translate matches a human translation more than ninety-five percent of the time is the day yet another job becomes obsolete thanks to AI. (Yes, I’m quite pessimistic about this. Why are we letting AI do all the fun stuff [art, writing, translation] and none of the soul-crushing or dangerous jobs?)

Advertisement: I made a wordpress blog for putting my BL Drama CD translations. If you happen to be a BL Drama CD fan, you can compare my manual translation to the one using this semi-automated workflow (ongoing, first disc completed).

[post-punk-revival]

Ok but real talk what the fuck is Musixmatch and is it just an AI that transcribes audio because some of these lyrics that end up on Genius right after the Musixmatch ones go up are not anything an actual person would write down after listening to the song

[ukrfeminism]

Submission! https://www.bbc.co.uk/news/uk-67123893

Victims of violent and sexual crimes are calling for court transcript costs to be cut after they were quoted "unaffordable" sums for them.

They told BBC Newsnight that charging thousands of pounds for copies of court hearings was "exploitative".

One rape survivor said she was quoted £7,500 for the transcript of her trial.

The Ministry of Justice (MoJ) said victims could ask a judge to order a transcript at public expense, but cases were not routinely transcribed.

"If the request is declined, the fee covers the considerable costs that come with writing up the audio recording of potentially weeks' worth of hearings," a MoJ spokesperson said.

Warning: This article contains details some readers may find distressing.

Juliana was raped by her former partner in 2020. 

"He drugged me. He recorded the video of himself raping me and he actually played it back to me after waking me up," she said. "He then threatened to send the video to my 88-year-old father. That's when I reported him to the police." 

Juliana's former partner was convicted by a jury at a trial, which lasted ten days.

She later wanted to revisit what had been said in court, but her request for a free copy of the transcript was rejected. 

She was told provisions were only made in exceptional circumstances, such as in murder and manslaughter cases, but those circumstances were "not met in your application".

Instead, she was advised to contact one of the companies outsourced by the government to supply transcripts. 

Acolad UK Limited quoted £7,459 for the transcription, Juliana said. The firm said the price was an estimate based on the length of the audio, which needed to be listened to by a transcriber.

"I just thought 'I can't afford this'. I had to stop working. My mental health was a mess," Juliana said. "Why do I have to pay for a service with data that is pertaining to me?"

The government's website says crown court hearings and those at civil and family courts are always recorded. Anybody can apply for a transcript of the proceedings.

It says victims will usually have to pay for the transcript, unless the court believes there are special circumstances. The final cost will depend on the size of the transcription, whether it's new or a copy, and other factors.

The court transcription service is outsourced to six companies in the UK, in a contract worth more than £17m.

BBC Newsnight found transcription costs at the six government-contracted firms varied from 80p per 72 words, to £1.71, for a 12-working-day transcription.

According to the government's guidance notes, Acolad UK Limited charges 80p per 72 words if the transcription is to be completed in 12 working days, which is listed as the cheapest turnaround option.

The company said pricing is based on the quantity of material to be transcribed, the level of urgency, and other factors.

"The sensitivity of the matter at hand - as in all legal and court proceedings - determines that use of AI-assisted tools is limited, and human expertise prioritised," Acolad said.

Crime victims have told BBC News that absorbing what is said in court can be incredibly difficult and traumatic, meaning they may have to rely on a transcription.

They argue having access to affordable transcripts allows them to go over the evidence and statements properly after cases have concluded.

According to the government, families in murder cases are entitled to free a copy of the judge's sentencing remarks following a conviction. 

'People deserve closure'

But Claire, whose ex-partner tried to kill her in 2020, said she was still quoted hundreds of pounds by Acolad for a transcription. 

"I was asleep and he cut my throat and then repeatedly stabbed me," Claire said. "I woke up to him trying to cut my neck."

Her former partner pleaded guilty to attempted murder and was sentenced to 18 years in prison. Because she had struggled to take in what had been said in the court hearing, Claire wanted to be able to read the judge's sentencing remarks.

She said she was not told a free copy could be requested from the court. Instead, she was advised to contact a transcription company. 

Claire said the firm quoted her almost £300 to provide a transcript of the judge's sentencing remarks. "I was quite shocked," she said. "I was homeless, I'm not working, I'm disabled, and I really need this for my closure - and I wasn't able to get it." 

She finally managed to get the transcript for free because somebody had already requested it and paid for the transcription work, which is standard procedure with all the companies.

"Some of the costs I've heard are astronomical, and these people deserve closure," Claire added. 

London's Victims' Commissioner, Claire Waxman said the current system must "urgently change". 

"Victims must be able to access accurate and timely transcripts, at no cost to themselves, to support their understanding and recovery, which is an essential part of their justice journey."

The Ministry of Justice said it was "incredibly rare for a victim to request a transcript of an entire trial" and it was more common for people to request the judge's sentencing remarks, which summarise the case against the defendant made at trial. It said that typically costs about £40.

If you've been affected by issues raised in this story, there is information and support available on BBC Action Line.

[remyfire]

[UPDATE, July 27th 2024] You all helped me meet my goal in less than 24 hours again and I am absolutely overwhelmed by your compassion and care!! I'm disabling reblogs so it doesn't keep circulating because for now, everything is taken care of ;v; Thank you so, so much

I was really not wanting to have to do this again so soon, especially while I've been spending time away from the Internet due to my poor mental health, but here we are. Peach is my wonderful, beautiful, perfect, sassy, cuddly, loving cat. She's 12 years old and exactly a month ago, we found out that she is diabetic. I love her so, so much and I will do anything I can to keep her healthy and give her a wonderful life as she enters her senior years, but unfortunately the costs of caring for a diabetic cat are more than I can afford right now.

Thanks to the AI boom, the transcription industry is drying up because if there's one thing these things are good at, it's transcribing audio files. Work has been increasingly difficult to come by, and while I've been fortunate to receive work this weekend, I won't be getting a paycheck for it for a couple of weeks, and even when I do, it absolutely has to go to my bills first and foremost and there will be no extra to put toward her cost of care. I continue to be on the hunt for jobs in my area and even had my first interview yesterday, but I haven't been lucky enough to receive an interview from any of the other places I've applied to thus far. And wouldn't you know it, but my credit card is just a hair away from maxing out.

Here's the current damage, which includes Peach's prescription food ($70), syringes ($20), and her most recent vet visit ($282), where they took her blood pressure and applied a glucometer in order to track her blood sugar levels for the next two weeks. Cats often have their diabetes go into remission, and if that happened then we would only need to maintain her diet—still expensive but far more affordable—but to get there, we need to see if this treatment is working, and that's going to take time.

I've set up a Ko-Fi goal here (a flat $400 in order to cover the PayPal fees) so you can see how much I still need to pay for her cost of care. I know that this is a miserable time to ask for donations and that we are so fatigued on giving already, but I legitimately have no choice. Until I can find work somewhere, I desperately need help.

Reblogs are appreciated. And here's one more cute photo of her in her little shirt from the vet as thanks for taking the time to read this post!!

[meret118]

But Whisper has a major flaw: It is prone to making up chunks of text or even entire sentences, according to interviews with more than a dozen software engineers, developers and academic researchers. Those experts said some of the invented text — known in the industry as hallucinations — can include racial commentary, violent rhetoric and even imagined medical treatments.

Experts said that such fabrications are problematic because Whisper is being used in a slew of industries worldwide to translate and transcribe interviews, generate text in popular consumer technologies and create subtitles for videos.

More concerning, they said, is a rush by medical centers to utilize Whisper-based tools to transcribe patients’ consultations with doctors, despite OpenAI’ s warnings that the tool should not be used in “high-risk domains.

The full extent of the problem is difficult to discern, but researchers and engineers said they frequently have come across Whisper’s hallucinations in their work. A University of Michigan researcher conducting a study of public meetings, for example, said he found hallucinations in eight out of every 10 audio transcriptions he inspected, before he started trying to improve the model.

A machine learning engineer said he initially discovered hallucinations in about half of the over 100 hours of Whisper transcriptions he analyzed. A third developer said he found hallucinations in nearly every one of the 26,000 transcripts he created with Whisper.The problems persist even in well-recorded, short audio samples. A recent study by computer scientists uncovered 187 hallucinations in more than 13,000 clear audio snippets they examined.

That trend would lead to tens of thousands of faulty transcriptions over millions of recordings, researchers said.

------

Yes, that does seem like a problem. If only there were a solution, such as, say, NOT USING IT!