This weekend in Generative Media

Penguin's copyright page forbids AI; Midjourney plans web AI editor; An AI feature film

News

• Penguin Random House books now explicitly say ‘no’ to AI training (The Verge)

• Midjourney plans to let anyone on the web edit images with AI (TechCrunch)

• ‘Where The Robots Grow’ Is AI’s First Feature (Forbes)

• Ted Sarandos Says After “Lots Of Hype” About AI, Key Question Remains: “Can It Help Make Better Shows And Films?” (Deadline)

• The AI opt-out models Meta, Musk's X, and the UK gov are proposing are simply not a good enough way for us to protect ourselves from data scraping (PC Gamer)

• This Prompt Can Make an AI Chatbot Identify and Extract Personal Details From Your Chats (Wired)

Software

• Hunch is the workspace for AI-first teams.

• Cartwheel is a powerful text-to-animation platform. Animate characters for your video, game, app, ad, social, AR or VR project simply by typing.

• Visual AI Backend Builder (Each Labs)

• AI Workflows (Coverr)

• Tora: Trajectory-oriented Diffusion Transformer for Video Generation (GitHub)

• EfficientViT: Multi-Scale Linear Attention for High-Resolution Dense Prediction (GitHub)

Research

• Look Ma, no markers: Holistic performance capture without the hassle (SIGGRAPH Asia 2024, project page)

• One-Step Diffusion via Shortcut Models (project page)

• UniCon: A Simple Approach to Unifying Diffusion-based Conditional Generation (project page)

• FlexGen: Flexible Multi-View Generation from Text and Image Inputs (project page)

• ControlMM: Controllable Masked Motion Generation (project page)

• DriveDreamer4D: World Models Are Effective Data Machines for 4D Driving Scene Representation (project page)

• SGEdit: Bridging LLM with Text2Image Generative Model for Scene Graph-based Image Editing (project page)

• MCGS: Multiview Consistency Enhancement for Sparse-View 3D Gaussian Radiance Fields (arXiv)

• Efficient Diffusion Models: A Comprehensive Survey from Principles to Practices (arXiv)

• Few-shot Novel View Synthesis using Depth Aware 3D Gaussian Splatting (arXiv)

• GlossyGS: Inverse Rendering of Glossy Objects with 3D Gaussian Splatting (arXiv)

• RoCoTex: A Robust Method for Consistent Texture Synthesis with Diffusion Models (arXiv)

Misc

• Tutorial: Rectified Flow (Qiang Liu, University of Texas Austin)

• gpt-4o-audio-preview audio input is currently 1,066 times more expensive than Google Gemini 1.5 Flash 8B audio input! (X)

• ChatGPT can now create Mind Maps. Here’s how to do it for free in a few seconds (X)

• I can confirm that with system prompt engineering and a high temperature, OpenAI's new gpt-4o-audio-preview model can be instructed to generate voices and any vocal style. (X)

• Jailbroken, R-rated NotebookLM is definitely something different. But it's interesting to listen to. (X) [NSFW language]