This weekend in Generative Media

GenAI disinfo is real; AI Herzog murder mystery doc; ESPN's AI avatar; Coke's AI ad

News

• Generative disinfo is real — you’re just not the target, warns deepfake tracking nonprofit (TechCrunch)

• AI Trained in Werner Herzog’s Works Wrote Murder Mystery Doc ‘About a Hero,’ Starring Vicky Krieps (Hollywood Reporter)

• ESPN is testing a generative AI avatar called ‘FACTS’ (The Verge)

• The Wall Street Journal is testing AI article summaries (The Verge)

• Behind the Scenes of Coca-Cola’s AI Reinvention of a Holiday Classic (AdWeek)

  • AI Coke Ad Shows We’ll Argue About Advertisements While the World Burns (Gizmodo)

• Bluesky says it won’t train AI on your posts (The Verge)

• Some of Substack’s Biggest Newsletters Rely on AI Writing Tools (Wired) [Since I mostly publish links, I don’t!]

• EU AI Act: Draft guidance for general purpose AIs shows first steps for Big AI to comply (TechCrunch)

• This 'lifelike' AI granny is infuriating phone scammers. Here's how - and why (ZDNet)

• Why It Matters That Google’s AI Gemini Chatbot Made Death Threats to a Grad Student (Inc.)

• Ars Live: Our first encounter with manipulative AI (Ars Technica)

Hardware and Software

• Generate, transform, and scale content easily using TikTok's AI-powered Symphony Creative Studio (TikTok blog)

• Google will let you make AI clip art for your documents (The Verge)

• This New Device Establishes Content Authenticity Using Any Digital Camera (PetaPixel)

• Region-Aware Text-to-Image Generation via Hard Binding and Soft Refinement (GitHub)

• DimensionX Create Any 3D and 4D Scenes from a Single Image with Controllable Video Diffusion (HuggingFace)

• Ultravox: An open-weight alternative to GPT-4o Realtime (ultravox.ai)

  • Talk to Ultravox Llama 3.1 8b (Powered by WebRTC ⚡️) (HuggingFace)

Research

• LLM2CLIP: Powerful Language Model Unlocks Richer Visual Representation (project page)

• Don't Look Twice: Faster Video Transformers with Run-Length Tokenization (project page)

• ProEdit: Simple Progression is All You Need for High-Quality 3D Scene Editing (project page)

• Scaling Mesh Generation via Compressive Tokenization (project page)

• MVDrag3D: Drag-based Creative 3D Editing via Multi-view Generation-Reconstruction Priors (project page)

• Self-Ensembling Gaussian Splatting for Few-shot Novel View Synthesis (arXiv)

• CameraHMR: Aligning People with Perspective (arXiv)

Misc

• How I used AI to design brand-aligned illustrations (Moty Weiss on Medium)

• CogVideoX 1.5 image2video with DimensionX orbit LoRA can really be something! (X)

• Brief tutorial on my prompting technique (X)

• „aI wIlL nEvEr rEpLaCe hUmaNs iN mOvIes aNd aDs“ CoCa Cola: hold my coke (X)