Today in Generative Media
AI music training costs; AI trained on YouTube risks privacy; AI in IATSE's deal
News
Training AI music models is about to get very expensive (MIT Technology Review)
AI companies train language models on YouTube’s archive − making family-and-friends videos a privacy risk (The Conversation)
Will AI Protections in IATSE’s Deal Be Enough to Calm Fears About Job Loss? (Variety)
Software
Meta starts testing user-created AI chatbots on Instagram (TechCrunch)
Character.AI now allows users to talk with AI avatars over calls (TechCrunch)
Is the World Ready for a Fully AI-based Complete Editing Suite? (nofilmschool)
Introducing AuraSR - An open reproduction of the GigaGAN Upscaler (fal) Demo on HuggingFace.
We’re extending our recent open-source project, StreamingT2V, to support High-Resolution Long Image-to-Video Generation (X) Code on GitHub.
Hunyuan-DiT v1.1 is now on Replicate. You can prompt it in English. (X) API on Replicate.
Thrilled to introduce nvTorchCam, our new PyTorch library designed to support the development of models using camera geometry like plane-sweep volumes (PSV) and related concepts like sphere-sweep volumes or epipolar attention, in a camera model-agnostic way! (X) Code on GitHub.
Research
Second-Hand Hallucinations: Investigating Perplexity's AI-Generated Sources (GPTZero)
3D Gaussian Splatting with Deferred Reflection (project page)
Misc
SHITHOLE-ON-SEA (YouTube) [Content warning: This 10-minute black comedy created with Midjourney, Pika and Bing isn’t for everyone. It’s got foul language, gore, body horror, etc. But, it’s also got dark humor, hints of plot and characters, and a coherent creative vision!]
Tried to get her to drink a pint while up there. Wait for it… (X)
It just keeps getting better (Reddit)