Synthetic Media & Generative AI News and Discussions

[firestar464]

[wjfox]

'Chubby filter' pulled from TikTok after user backlash

21 March 2025

A viral filter which made people appear overweight has been removed from TikTok, after the BBC reported it had sparked a wave of user criticism.

Known as a "chubby filter", the artificial intelligence (AI) tool took a photo of a person and edited their appearance to look as though they had put on weight.

Many people have shared their "before and after" images on the platform with jokes about how different they looked - however, others said it was a form of "body shaming" and should not be permitted.

Experts have also warned the filter could fuel a "toxic diet culture" online and potentially contribute to eating disorders.

TikTok said the filter had been uploaded by a CapCut, which is separate to TikTok but has the same parent company, ByteDance.

https://www.bbc.co.uk/news/articles/c4gmvjvv6vjo

[wjfox]

[firestar464]

Try it at https://preview.reve.art/

[wjfox]

ChatGPT’s image-generation feature gets an upgrade

11:32 AM PDT · March 25, 2025

During a livestream on Tuesday, OpenAI CEO Sam Altman announced the first major upgrade to ChatGPT’s image-generation capabilities in over a year.

ChatGPT can now leverage the company’s GPT-4o model to natively create and modify images and photos. GPT-4o has long underpinned the AI-powered chatbot platform, but until now, the model only been able to generate and edit text — not images.

Altman said GPT-4o native image generation is live today in ChatGPT and Sora, OpenAI’s AI video generation product, for subscribers to the company’s $200-a-month Pro plan. OpenAI says the feature is rolling out soon to Plus and free users of ChatGPT, as well as to developers using the company’s API service.

GPT-4o with image output “thinks” a bit longer than the image generation model it effectively replaces, DALL-E 3, to make what OpenAI describes as more accurate and detailed images. GPT-4o can edit existing images, including images with people in them — transforming them or “inpainting” details like foreground and background objects.

OpenAI didn’t reveal which image data it used to enable the new image generation capabilities. Many generative AI vendors see training data as a competitive advantage and so keep it and any information related to it close to the chest. But training data details are also a potential source of IP-related lawsuits, another disincentive for companies to reveal much.

https://techcrunch.com/2025/03/25/chatg ... n-upgrade/

[firestar464]

Introducing 4o Image Generation

https://openai.com/index/introducing-4o ... eneration/

Istg today is shipping day. Deepseek, some genAI startup, Google, and now OAI

[wjfox]

Krisp is using AI to help Indians sound like Americans on calls

March 26, 2025

Audio startup Krisp on Wednesday said it is launching a new feature that uses AI to change a user’s accent during calls. The company is initially rolling out support for changing Indian English accents to U.S. English.

The startup says the accent conversion process preserves the speaker’s voice and only switches phonemes to match American accents. The feature has apparently been tested in enterprise environments, and a beta version is now coming to the Krisp desktop app. Users can turn the feature on any time during or before calls.

Arto Minasyan, the company’s co-founder, said the idea for the feature stemmed from a problem he faced in his conversations. “Many people don’t understand my accent even though I am speaking English well. We thought changing accents might help people understand each other much better. We started working on this problem two years ago and now we are releasing it in beta,” Minasyan said.

However, when this reporter tested the feature, the processed voice didn’t sound natural and even missed some words at times. The company attributed those faults to this being a beta release, saying the model would improve over time.

Krisp claims that when it tested the feature with enterprises, sales conversion rates rose by 26.1% and revenue per book jumped by 14.8%.

https://techcrunch.com/2025/03/26/krisp ... ng-a-call/

[weatheriscool]

[firestar464]

I mean this does infringe on personality rights if it actually mimics the style, which would be both illegal (in California IIRC) and unethical. (Parody is a valid exception if I'm not wrong.) Thus I am glad that OpenAI has committed to not mimicking the style of of living artists, though evidently their safeguards are not effective enough, given that people are still successfully creating stuff in his style.

Also if you carefully watch the video where Miyazaki criticizes the tech, while yes, he is unhappy with the tech, his anger is directed at the trivialization of gore and disability. His reason for opposition is the idea that people are losing faith in their ability to draw.

I don't doubt human ability at all. (Who do you think has been producing art all this time?) I'm just of the belief that a machine could help create images more efficiently.

Video in question:

[Cyber_Rebel]

It's also worth noting that Miyazaki's comment was literally almost a decade ago, and I doubt he could've known where we'd be right about now. The level of detail and prose within 4o's image generation capabilities is literally another watershed moment, to the point I think people will look back to right around now when they wonder where we crossed the Rubicon. It's advantageous for GPT to learn from these styles of well-known artist, but it won't be for long before people creatively devise their own styles, we couldn't have imagined prior to this technology.

Can you truly imagine the internet, media, and your sense of perception towards it remaining the same 5 years out from now? Hell, make that 2-3 at best, considering where we were a mere 2-3 years ago. The moment videos become truly longer, sustainable, and reach a mass critical point like what happened just now? You can basically isekai yourself to any fiction you'd likely be able to dream of.

I'm not saying that it's over for artist or that humans won't still be within the mix of things for a while. But the old ways are leaving. It's certainly better within this new paradigm to develop or redevelop critical thinking skills.

[firestar464]

I mean, AlphaGo learned from Go players not simply to mimic them; it ultimately did so to do things they never did before.

[wjfox]

[wjfox]

[wjfox]

[wjfox]

[firestar464]

[Yuli Ban]

Yeah there's a part of me that always felt skeevy about the fact that Miyazaki's style in particular is what OpenAI focused on. If that was meant as a form of respect to Miyazaki like DALL-E was for Dali, that backfired extraordinarily. Virtually no one other than AI art users see it as anything resembling respect, especially considering Miyazaki's infamous take (some say it was misrepresented, but let's be clear here, the man was hostile towards computer/digital art, only the most copium-huffers would ever think he feels positively towards AI generated art. Maybe, maybe, maybe very very very long shot the old man can find something curious about how a computer sees the world, but he'd undoubtedly hate it being used to replace artists)
It's like, people wonder why I largely stopped advertising synthetic media and took a massive step back from talking or posting about it publicly. Just look at AI artist behavior to see why! I mean f*ck, the latest controversy is Donald Trump posting an AI image of himself as the Pope. Meanwhile, researchers literally just socially engineered Redditors with masses of undisclosed AI chatbots. Like that's the kind of bullshit people are going to associate alongside the likes of the Stanford Prison Experiment in terms of how all kinds of unethical it is.

That's what people currently associate generative AI with: the worst people you know doing the shittiest things you can imagine.

[firestar464]

Agree with almost everything. Just very slight nitpick:

but let's be clear here, the man was hostile towards computer/digital art

I mean, Princess Mononoke was CGI-assisted. He isn't altogether opposed to it; just indeed, he wouldn't replace anyone with it, you know?

[weatheriscool]

Hybrid AI model crafts smooth, high-quality videos in seconds
https://techxplore.com/news/2025-05-hyb ... -high.html
by Alex Shipps, Massachusetts Institute of Technology

What would a behind-the-scenes look at a video generated by an artificial intelligence model be like? You might think the process is similar to stop-motion animation, where many images are created and stitched together, but that's not quite the case for "diffusion models" like OpenAI's SORA and Google's VEO 2.

Instead of producing a video frame-by-frame (or "autoregressively"), these systems process the entire sequence at once. The resulting clip is often photorealistic, but the process is slow and doesn't allow for on-the-fly changes.

Scientists from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) and Adobe Research have now developed a hybrid approach, called "CausVid," to create videos in seconds. Much like a quick-witted student learning from a well-versed teacher, a full-sequence diffusion model trains an autoregressive system to swiftly predict the next frame while ensuring high quality and consistency. CausVid's student model can then generate clips from a simple text prompt, turning a photo into a moving scene, extending a video, or altering its creations with new inputs mid-generation.

[firestar464]

reposted by Jimmy