Today in Generative Media

California's AI replica law; Ian Holm's return in 'Alien: Romulus'; NaNoWriMo defends AI

News

• California Passes Law Requiring Consent for AI Digital Replicas of Dead Performers (Variety)

• How Ian Holm’s Return in ‘Alien: Romulus’ Brings SAG-AFTRA’s Digital Replicas Work Into Focus (Variety)

• NaNoWriMo is in disarray after organizers defend AI writing tools (The Verge)

• Dutch regulator slaps Clearview AI with $33 million fine and threatens executive liability (The Verge)

• X users play Elon Musk at his own game after he posts AI image of Harris as communist dictator (The Independent)

• AI-Fakes Detection Is Failing Voters in the Global South (Wired)

• Canva says its AI features are worth the 300 percent price increase (The Verge)

Opinion

• a16z’s Joshua Lu says AI is already radically changing video games and Discord is the future (TechCrunch)

• Using The Machine: Evolving Notions Of Originality In Copyright Law And AI-Assisted Art (Mondaq)

• Written before the current wave of generative AI, this is an interesting historically-informed view of whether computers can create “Art.” It deals thoughtfully with the difference between AI as tool and AI as artist, and how definitions shift over time. (X)

Software

• Camera Motion is here in Dream Machine 1.6! (X) Luma AI debuts Dream Machine 1.6 video generator with camera motions (VentureBeat)

• WhisperX: fast automatic speech recognition (70x realtime with large-v2) with word-level timestamps and speaker diarization (GitHub)

• Update on Advanced LivePortrait (Video to Video) - Faster and real-time preview - Create a video without a video - Edit facial expressions in photos - Insert facial expressions into videos - Create animations using multiple facial expressions - Extract facial expressions from sample photos. (X)

• Cube now supports an experimental AI model that generates retopologized meshes, which are cleaner and use ~98% fewer triangles (X)

• Mini-Omni is an open-source multimodel large language model that can hear, talk while thinking. Featuring real-time end-to-end speech input and streaming audio output conversational capabilities (HuggingFace)

• TorchGeo is a python library for deep learning with geospatial datasets. This large update contains new datasets, foundation models, and more (X)

• Official 🤗 Gradio demo for CSGO: Content-Style Composition in Text-to-Image Generation. (HuggingFace)

Research

• Improving 2D Feature Representations by 3D-Aware Fine-Tuning (project page)

• VQ4DiT: Efficient Post-Training Vector Quantization for Diffusion Transformers (arXiv)

• OD-VAE: An Omni-dimensional Video Compressor for Improving Latent Video Diffusion Model (arXiv)

• DynOMo: Online Point Tracking by Dynamic Online Monocular Gaussian Reconstruction (arXiv)

Misc

• ACM -- disclose every use of LLMs everywhere, even when you use an LLM to generate archaic latex to make your table. ML Papers -- we used LLMs to simulate human responses, to finetune an LLM, and then evaluated them using another LLM (X)