Today in Generative Media

Newsom signs AI likeness bills; Challenging genAI myths; Meta training AI on UK posts

News

• California Gov. Gavin Newsom Signs AI Bills That Give Performers Greater Protections & Control Over Their Likenesses, Limit Use Of Election Deepfakes — Update (Deadline)

• Challenging The Myths of Generative AI (Tech Policy Press)

• Meta to push on with plan to use UK Facebook and Instagram posts to train AI (The Guardian)

• Google will begin flagging AI-generated images in Search later this year (TechCrunch)

• Snap is introducing an AI video-generation tool for creators (TechCrunch)

• Snap’s new AI feature lets you create Snapchat Lenses by simply describing them (TechCrunch)

• EA execs say generative AI is "not merely a buzzword for us, it's the very core of our business," then pretend to tell a computer to generate buildings live on stage (gamesradar.com)

• AI datacenters more than 600 percent worse for environment than tech companies claimed. (Futurism)

Software

• Discover DeepEditor, the film editing tool that makes filmmaking agile. Refine dialogue, enhance performances and reduce shoot time. Perfect your story without returning to set. (Flawless AI)

• Build and scale creative products with the world's most popular and intuitive video generation models in the Dream Machine API (Luma Labs) Demo on HuggingFace (requires API Key)

• Make anything, anywhere. With the Runway API (RunwayML)

• GeoCalib 📸 Single-image Calibration with Geometric Optimization (GitHub)

• We’re offering Rapid Flux Training – LoRAs trained in under 5 minutes! (CivitAI)

• Slang.D Gaussian Splatting Rasterizer (GitHub) This allows for a single code-base of the rendering code which has the capability to run at the same time under the differentiable framework of your choice (i.e Pytorch) and under the actual renderer that can be released for production (i.e Vulkan). This makes significant strides toward maintainability and reduces the likelihood of errors.

Research

• NPGA: Neural Parametric Gaussian Avatars (SIGGRAPH Asia 2024, project page)

• InstantDrag: Improving Interactivity in Drag-based Image Editing (SIGGRAPH Asia 2024, project page)

• MM3DGS SLAM: Multi-modal 3D Gaussian Splatting for SLAM Using Vision, Depth, and Inertial Measurements (IROS 2024, project page)

• WonderWorld: Interactive 3D Scene Generation from a Single Image (project page)

• Seed-Music (project page) Our unified framework enables the generation of music with expressive vocals in multiple languages, allows precise note-level adjustments to model outputs, and offers the ability to integrate users' own voices into music creation, and much more.

• DrawingSpinUp: 3D Animation from Single Character Drawings (project page)

• PrimeDepth: Efficient Monocular Depth Estimation with a Stable Diffusion Preimage (arXiv)

• CSS: Overcoming Pose and Scene Challenges in Crowd-Sourced 3D Gaussian Splatting (arXiv)

Misc

• DOGSTUDIO/DEPT® collaborated with The Coca-Cola Company on a film called "Neon Nights" which I directed for the 2024 Runway AI Film Festival. (LinkedIn) The goal was to see how far we could go leveraging a maximum of AI technologies in the smartest way, but also not at the expense of a good story. Today, here is the full-film and a complete Making Of playing just behind it.

• Google's Notebook LM takes whatever docs, pdf or links you give and turns it into a podcast! HeyGen can easily animate that into a video presentation. Aside from editing, this is 100% AI generated.... (X)

• Guillermo Del Toro on AI, emotion and the value of art. (X)

• Using photogrammetry of an old TV's circuitry to generate an aerial view of a post-apocalyptic cityscape through RunwayML Gen-3 V2V (X)

• You can animate these beautiful ink-style transitions with Runway and Midjourney. (X)