This weekend in Generative Media

Studios adopt genAI slowly; AI finds more Nazca lines; Report from AI film screening

News and Opinion

• Large studios will likely take their time adopting generative AI for content creation. Social media isn’t hesitating. (Deloitte)

• Hundreds More Nazca Lines Emerge in Peru’s Desert (New York Times) With drones and A.I., researchers managed to double the number of mysterious geoglyphs in a matter of months.

• One big room full of hopes and schemes: An anecdotal report of a major AI film screening and the reception that followed. (Mike Gioia on Substack)

• Labelers training AI say they're overworked, underpaid and exploited by big American tech companies (CBS News)

• 'It might pave the way for novel forms of artistic expression': Generative AI isn't a threat to artists — it's an opportunity to redefine art itself (LiveScience)

• Moonvalley wants to build more ethical video models (TechCrunch)

• Coca Cola’s AI-Generated Ad Controversy, Explained (Forbes)

• Want to speak Italian? Microsoft AI can make it sound like you do. (Washington Post)

• HarperCollins strikes AI training deal with unnamed company amid rising copyright tensions between publishers and AI firms (Fortune)

• ChatGPT’s Poetry is Incompetent and Banal (Ernest Davis, NYU) [PDF]

Software

• Expand Video allows you to seamlessly generate beyond the edges of a frame to turn horizontal videos into vertical videos, and vice versa - without compromising details or framing (RunwayML)

  • This is the real breakthrough of Runway ML new Expand feature.. make ANY movie scene « compatible » for a mobile screen format. (X)

• Introducing Frames: An image generation model offering unprecedented stylistic control (RunwayML)

• Introducing v4: Better audio, sharper lyrics, and more dynamic song structures now available in beta to Pro & Premier (Suno)

• Playing around with Hotshot’s new video reference (Vid2Vid). A lot of potential here and what's awesome is you can use a character reference (X)

• aiOla unveils open source AI audio transcription model that obscures sensitive info in realtime (VentureBeat)

• ComfyUI-LTXTricks A set of nodes that provide additional controls for the LTX Video model (GitHub)

• ComfyUI_AdvancedReduxControl (GitHub)

• Say hello to the all-new [Luma] DreamMachine.  Your visual thought partner—where ideas become reality. Ideate, visualize, and share your ideas with the world. Available now. (X)

Research

• 🎞️ MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views (NeurIPS 2024 project page)

• ARM: Appearance Reconstruction Model for Relightable 3D Generation (project page)

• Direct and Explicit 3D Generation from a Single Image (project page)

• Find Any Part in 3D (project page)

• Stylecodes: Encoding Stylistic Information For Image Generation (project page)

• Oscillation inversion: Understanding the structure of large flow models through the lens of inversion methods (project page)

• StableV2V: Stablizing Shape Consistency in Video-to-Video Editing (project page)

• 3D Convex Splatting: Radiance Field Rendering with 3D Smooth Convexes (project page)

• Beyond Gaussians: Fast and High-Fidelity 3D Splatting with Linear Kernels (arXiv)

• OminiControl: Minimal and Universal Control for Diffusion Transformer (arXiv)

Misc

• Build and design an AI image generator app for iOS without ANY design or coding experience (X)