This weekend in Generative Media

Adobe's AI video model is here; AI voice chat reactions; AI content moderation works

News and Opinion

• Adobe’s AI video model is here, and it’s already inside Premiere Pro (The Verge)

• Who Is That You Are Chatting With? Oh, Just ChatGPT. (New York Times)

• I just had a conversation with Meta AI Voice — and it’s way better than I expected (Tom’s Guide)

• Ahead of Black Ops 6’s Launch, Activision Says Call of Duty’s AI-Powered Voice Moderation Has Already Had a Massive Impact on Toxicity (IGN)

• ByteDance lays off hundreds of TikTok employees in shift to AI content moderation (TechCrunch)

• Photoshop is getting a bunch of new AI tools (The Verge)

• Zoom’s custom AI avatar tool may come with risks (TechCrunch)

• There's a future where you and your graphics card decide what a game looks like, not the developers (PC Gamer)

• Advertisers share what's working in AI and what's not as the hype cycle fades (Business Insider)

• Are Tech's Generative AI "Fair Use" Dominoes Starting to Fall? (Peter Csathy on Substack)

• LLMs don’t do formal reasoning - and that is a HUGE problem (Gary Marcus on Substack)

• ‘Dune 2’ Director Denis Villeneuve Says He’ll Miss the ‘Collective Act of Creativity’ if ‘One Day’ Films Are Made ‘Just With a Computer’ (Variety)

Software

• How to Scale Video Editing With an AI Storytelling Partner (nofilmschool)

• Towards Natural Image Matting in the Wild via Real-Scenario Prior (GitHub)

• LivePortrait for Nuke (GitHub)

Research

• Generalizable and Animatable Gaussian Head Avatar (NeurIPS 2024, project page)

• Progressive Autoregressive Video Diffusion Models (project page)

• ViBiDSampler: Enhancing Video Interpolation Using Bidirectional Diffusion Sampler (project page)

• OmniBooth: Learning Latent Control for Image Synthesis with Multi-modal Instruction (project page)

• FLUX-IR: Learning Efficient and Effective Trajectories for Differential Equation-based Image Restoration (project page)

• Generative Portrait Shadow Removal (SIGGRAPH Asia 2024, arXiv)

• DART: Denoising Autoregressive Transformer for Scalable Text-to-Image Generation (arXiv)

• Poison-splat: Computation Cost Attack on 3D Gaussian Splatting (arXiv)

• StoryNavi: On-Demand Narrative-Driven Reconstruction of Video Play With Generative AI (arXiv)

Misc

• There is an opportunity for a new distribution channel to emerge because interactive video will probably require interactivity features that YouTube / TikTok / etc. don't offer to either creators or consumers. (X)

• Somedays I just love AI so much (X)

• UV Unwrapped Head LoRA (Flux.1 D) (X)