Today in Generative Media

AI at art school; Russia seeds AI with disinfo; Sarandos: AI to make movies 10% better

News and Opinion

• An AI is going to art school — and might earn a diploma. Meet Flynn. (Washington Post)

• Russia seeds chatbots with lies. Any bad actor could game AI the same way. (Washington Post)

• Ted Sarandos Responds To James Cameron’s Vision Of AI Making Movies Cheaper: “There’s An Even Bigger Opportunity To Make Movies 10% Better” (Deadline)

• AI-generated music accounts for 18% of all tracks uploaded to Deezer (Reuters)

• Adobe Gets Bullied Off Bluesky (Futurism)

• Say Goodbye to Your Kid’s Imaginary Friend (New York Times)

• Creators Are Losing the AI Copyright Battle. We Have to Keep Fighting (Guest Column) (Hollywood Reporter)

• The real argument artists should be making against AI (Vox)

• Capsule captures $12M to build the next version of its AI video editor for brands (TechCrunch)

Software

• Ponder AI can learn the style of filmmakers, analyse your footages, edit the full video all on its own and.. you can even edit the sequence using text prompt (X)

  • Tim Tan and I spent the last few weeks building Ponder Studio to edit a vlog cuz my Adobe subscription expired (X)

• Google adds Veo 2 video generation to Gemini app (Ars Technica)

• Whisk Animate is now available globally to Google One AI premium subscribers! With Veo 2, turn your Whisk-generated images into vivid eight-second animated clips - perfect for reimagining your images in different styles or bringing new ideas to life. (X)

• You can now train FLUX LoRAs in 30 seconds! (X)

• This is the official implementation of the paper "Volumetrically Consistent 3D Gaussian Rasterization" (GitHub)

Research

• Seaweed: A video generation foundational model (project page)

• Liquid: Language Models are Scalable and Unified Multi-modal Generators (project page)

• Packing Input Frame Context in Next-Frame Prediction Models for Video Generation (project page)

• SHeaP: Self-Supervised Head Geometry Predictor Learned via 2D Gaussians (project page)

• NormalCrafter: Learning Temporally Consistent Normals from Video Diffusion Priors (project page)

• VideoGameBench (project page)

• DNF-Avatar: Distilling Neural Fields for Real-time Animatable Avatar Relighting (project page)

• St4RTrack: Simultaneous 4D Reconstruction and Tracking in the World (project page)

• Second-order Optimization of Gaussian Splats with Importance Sampling (project page)

• Art3D: Training-Free 3D Generation from Flat-Colored Illustration (project page)

• FRPGS: Fast, Robust, and Photorealistic Monocular Dynamic Scene Reconstruction with Deformable 3D Gaussians (IEEEXplore)

• LightHeadEd: Relightable & Editable Head Avatars from a Smartphone (arXiv)

Misc

• Kling AI now can swap out any actor from any film scene. fast shots, complex motion, doesn’t matter. (X)

• AI Channels are Infecting Youtube (YouTube)