Summary
This weekly AI news roundup from AI Search covers a packed set of releases across video, image, 3D, robotics, and language models. Headline items include Google’s Gemini Omni 1.2 video model, OpenAI’s latest image model that can produce animated GIFs, NVIDIA’s open-source weather simulation AI, and the release of Kimi K3.1 alongside Tencent’s new Hy4 model.
A large part of the video focuses on open tools that can be run locally. Highlights include a model that generates full 3D hairstyles with movable joints from a single image (about 2.45 GB, Apache 2 license), a 4x video upscaler that enhances only key frames with a large model and fills in the rest with a smaller one, and Nitro Fusion 2, a Z-Image-based generator that produces images in one or two steps. Another entry turns the open video model MiniMax H3 into an image generator and editor that accepts up to eight reference images and outputs up to 16 megapixels.
The episode also touches on the Open Robo Kit robotics toolkit, the real-time multimodal voice system Moss Live, new research on improving AI models in weight space without further training, and several additional 3D generators. Each segment points to code and model downloads for viewers who want to try them.
📺 Source: AI Search · Published November 08, 2026
🏷️ Format: Roundup
