Today in Generative Media

Artists leak Sora, OpenAI cuts access; IMAX embraces AI; New genre: AI nature slop

[I’ll be taking the rest of the week off for the US Thanksgiving Holiday. See you next week!]

News and Opinion

• OpenAI hits pause on video model Sora after artists leak access in protest (Washington Post)

• OpenAI Shuts Down Sora Access After Artists Released Video-Generation Tool in Protest: ‘We Are Not Your PR Puppets’ (Variety)

• Artists Stick It to ‘AI Overlords’ and Leak OpenAI’s Sora Video Generator (Gizmodo)

• IMAX embraces AI to expand original content reach (TechCrunch)

• A New Genre of Bad A.I. Art Takes the Stage: Nature Slop (ArtNet)

• Inflection AI CEO says it’s done trying to make next-generation AI models (TechCrunch)

Software

• FLUX fine-tunes are now fast (Replicate)

• Mochi 1 LoRA Fine-tuner (GitHub)

• GenChess: Make Your Next Move a Masterpiece (Google Labs)

• Ai2 releases new language models competitive with Meta’s Llama (TechCrunch)

• Google’s connecting Spotify to its Gemini AI assistant (The Verge)

• Anthropic says Claude AI can match your unique writing style (The Verge)

Research

• Generative Omnimatte: Learning to Decompose Video into Layers (project page)

• Video-Guided Foley Sound Generation with Multimodal Controls (project page)

• AC3D: Analyzing and Improving 3D Camera Control in Video Diffusion Transformers (project page)

• Towards High-Fidelity 3D Portrait Generation with Rich Details by Cross-View Prior-Aware Diffusion (project page)

• SelfSplat: Pose-Free and 3D Prior-Free Generalizable 3D Gaussian Splatting (arXiv)

• Gaussian Scenes: Pose-Free Sparse-View Scene Reconstruction using Depth-Enhanced Diffusion Priors (arXiv)

• DROID-Splat: Combining end-to-end SLAM with 3D Gaussian Splatting (arXiv)

• GSurf: 3D Reconstruction via Signed Distance Fields with Direct Gaussian Supervision (arXiv)