This weekend in Generative Media

'House of David' uses AI; Swedish film's AI visual dub; Meta tests genAI comments

News and Opinion

• ‘House of David’ Creator Explains Using AI to Create the Goliath Origin Sequence (EXCLUSIVE) (Variety)

  • With 22 million views in just 17 days, House of David is one of the first major productions to use generative AI at scale. (X)

• Swedish Film 'Watch the Skies' Set for US Release With AI 'Visual Dubbing' (Decrypt)

• Meta spotted testing AI-generated comments on Instagram (TechCrunch)

• Exploring the frontier of AI, robotics and creativity in content production (WPP) For the first time, Boston Dynamics' Atlas robot steps behind the camera – enhancing precision, expanding creative potential and keeping human operators out of hazardous environments. (YouTube)

• AI’s answers on China differ depending on the language, analysis finds (TechCrunch)

• CEO of AI ad-tech firm pledging “world free of fraud” sentenced for fraud (Ars Technica)

Software

• Google is rolling out Gemini’s real-time AI video features (The Verge)

• DeTikZify: Synthesizing Graphics Programs for Scientific Figures and Sketches with TikZ (HuggingFace)

• This repository contains the implementation of the gemma PyPI package. A JAX library to use and fine-tune Gemma. (GitHub)

• We are happy to announce that ComfyUI now natively supports Hunyuan3D 2.0 and Hunyuan3D 2.0 MV (Multi-View) model series! (X)

Research

• SplatVoxel: History-Aware Novel View Streaming without Temporal Training (project page)

• Lux Post Facto: Learning Portrait Performance Relighting with Conditional Video Diffusion and a Hybrid Dataset (project page)

• MotionStreamer: Streaming Motion Generation via Diffusion-based Autoregressive Model in Causal Latent Space (project page)

• DeepMesh: Auto-Regressive Artist-mesh Creation with Reinforcement Learning (project page)

• FlexWorld: Progressively Expanding 3D Scenes for Flexiable-View Synthesis (project page)

• Sonata: Self-Supervised Learning of Reliable Point Representations (project page)

• InfiniteYou: Flexible Photo Recrafting While Preserving Your Identity (project page)

• LHM: Large Animatable Human Reconstruction Model for Single Image to 3D in Seconds (project page)

• Dynamic Point Maps: A Versatile Representation for Dynamic 3D Reconstruction (project page)

• MagicMotion: Controllable Video Generation with Dense-to-Sparse Trajectory Guidance (project page)

• The Cybernetic Teammate (Ethan Mollick on Substack) The Cybernetic Teammate: A Field Experiment on Generative AI Reshaping Teamwork and Expertise (SSRN)

• Bézier Splatting for Fast and Differentiable Vector Graphics (arXiv)

• Unleashing Vecset Diffusion Model for Fast Shape Generation (arXiv)

• Unlock Pose Diversity: Accurate and Efficient Implicit Keypoint-based Spatiotemporal Diffusion for Audio-driven Talking Portrait (arXiv)