Today in Generative Media

Pocoyo's AI pipeline; WGA wants AI protections from CBS News; AI voice risks in court

News and Opinion

• Iconic Children’s Show ‘Pocoyo’ Has A New Owner — And A New AI-Powered Production Pipeline (Cartoon Brew)

• WGA Zeroes In On AI Protections In CBS News Contract Negotiations To “Safeguard Journalism As A Profession” (Deadline)

• AI-Generated Voice Evidence Poses Dangers in Court (Lawfare)

• ‘Everything Everywhere All at Once’ Effects Guru Talks Pretend VFX and the Coming AI “Tsunami” (Hollywood Reporter)

• How ProPublica Uses AI Responsibly in Its Investigations (ProPublica)

• Microsoft’s new Xbox Copilot will act as an AI gaming coach (The Verge)

• I cloned my voice in seconds using a free AI app, and we really need to talk about speech synthesis (TechRadar)

• OpenAI’s ‘creative writing’ AI evokes that annoying kid from high school fiction club (TechCrunch)

• OpenAI declares AI race “over” if training on copyrighted works isn’t fair use (Ars Technica)

• Google calls for weakened copyright and export rules in AI policy proposal (TechCrunch)

Software

• Mirage is LIVE: Generate energetic, high-converting ads with people that don’t exist — complete with animated body language and micro-expressions — using Mirage, the first foundation model built to generate UGC-style content. (X)

• BlenderMCP connects Blender to Claude AI through the Model Context Protocol (MCP), allowing Claude to directly interact with and control Blender. This integration enables prompt assisted 3D modeling, scene creation, and manipulation. (GitHub)

  • Here’s a demo of me creating a “low-poly dragon guarding treasure” scene in just a few sentences (X)

• Introducing Gen 2: The Next Frontier of Spatial Motion (move.ai)

• Open-Sora: Democratizing Efficient Video Production for All (GitHub) 11B model achieves on-par performance with 11B HunyuanVideo & 30B Step-Video on 📐VBench & 📊Human Preference. 🛠️ Fully open-source: checkpoints and training codes for training with only $200K.

• Sesame, the startup behind the viral virtual assistant Maya, releases its base AI model (TechCrunch)

• We used Mistral OCR with Claude 3.7 to create blog-style overviews for arXiv papers (X)

• AI-powered search inside research papers 🔥 (LinkedIn)

Research

• LBM: Latent Bridge Matching for Fast Image-to-Image Translation (project page)

• Sparse VideoGen: Accelerating Video Generation with Spatial-Temporal Sparse Attention by 2x with High Fidelity (project page)

• Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Models (project page)

• Multimodal Representation Alignment for Image Generation: Text-Image Interleaved Control Is Easier Than You Think (arXiv)

Misc

• Physics-inspired 3D Vision and Imaging (CVPR workshop)

• We put 4 different image generators to the test - Midjourney, Image FX, Photon by Luma Labs, and Mystic by Magnific - challenging them to recreate the signature looks of legendary filmmakers. (X)

• You don't need to know Blender or vibe code with Claude to do 3DtoAI (X)

• THIS is what we've been promised for so long! Multimodal image Editing! And Gemini 2.0 did it! Generate and image, and then ASK for a change without completely regenerating! (X)

• some cool examples with Gemini 2.0 native image output (X)