Today in Generative Media

OpenAI's Hollywood pitch; Amazon's AI copyright nightmare; AI in sci-fi must change

News and Opinion

• OpenAI takes its pitch to Hollywood creatives after launching controversial video tool (Los Angeles Times)

• Amazon is blundering into an AI copyright nightmare (The Verge)

• The Electric State Confirms How Sci-Fi Stories Use A.I. Characters Must Change (Den of Geek)

• Bob Iger Says AI May Be “Most Powerful Technology That Our Company Has Ever Seen” (Hollywood Reporter)

• Andor creator refuses to publish scripts thanks to AI (The Verge)

• Study finds AI-generated meme captions funnier than human ones on average (Ars Technica)

• Academics accuse AI startups of co-opting peer review for publicity (TechCrunch)

• fxpodcast: Oscar-winning legend Rob Legato on Stability AI & the future of VFX (FXGuide)

Software

• A Unity package that enables seamless communication between Unity and Large Language Models (LLMs) like Claude Desktop via the Model Context Protocol (MCP). (GitHub)

• AbletonMCP connects Ableton Live to Claude AI through the Model Context Protocol (MCP), allowing Claude to directly interact with and control Ableton Live. (GitHub)

• Larger. Sharper. More real. Upscale and enhance any image 16x. (Topaz Labs)

• StarVector is a multimodal LLM for Scalable Vector Graphics (SVG) generation, producing structured SVG code directly from images and text. (HuggingFace)

• This is an official demo for the paper "GenStereo: Towards Open-World Generation of Stereo Images and Unsupervised Matching". Given an arbitrary reference image, GenStereo can generate the corresponding right-view image. (HuggingFace)

• NVIDIA Open Sources 3D Gaussian Ray Tracing and 3D Gaussian Unscented Transform (Radiance Fields)

• Cosmos-Transfer1 is a pre-trained, diffusion-based conditional world model designed for multimodal, controllable world generation. It creates world simulations based on multiple spatial control inputs across various modalities, such as segmentation, depth, and edge maps. (GitHub)

• This project is a node-based implementation for video generation using the Wan2.1 model, with a focus on start and end frame guidance (GitHub)

• Introducing Natural Language Editing—just ask for what you want. Free to try for a limited time in FLORA. Powered by @GoogleDeepMind ’s Gemini 2.0 Flash. (X)

Research

• Bolt3D: Generating 3D Scenes in Seconds (project page)

• StarVector: Generating Scalable Vector Graphics Code From Images And Text (project page)

• CameraCtrl II: Dynamic Scene Exploration via Camera-controlled Video Diffusion Models (project page)

• AudioX: Diffusion Transformer for Anything-to-Audio Generation (project page)

• MVGSR: Multi-View Consistency Gaussian Splatting for Robust Surface Reconstruction (project page)

• Impossible Videos (project page)

• MeshPad: Interactive Sketch-Conditioned Artist-Designed Mesh Generation and Editing (project page)

• Cube: A Roblox View of 3D Intelligence (arXiv)

• Multi-view Reconstruction via SfM-guided Monocular Depth Estimation (arXiv)

Misc

• Google Gemini can analyze videos to create extra frames and shots. (X)

• UV-aware texture generation powered by Gemini Flash 2.0 Experimental. Now texture re-imagination is just an LLM call away. Crazy (X)

• I'm excited about the recent feature we've been working on at Pika Labs - action editing! (X)

• Moving Memories with [Luma] Ray2 Img-to-Vid. Restore your old photo albums by turning them into videos—bringing them back to life. (X)

• Initial version of depth control lora for Wan2.1 1.3b (X)

• Wan2.1 loras trained on Replicate now support image-to-video. So you can have your trained AI character deliver coffee to someone, if you like. (X)