Summary
Matt Wolfe’s weekly deep dive rounds up 14 AI updates from the past week, with hands-on tests of several of the biggest releases.
The episode opens with OpenAI’s new image model, Images 2.6, now in ChatGPT and the API. Wolfe tests its headline feature, animated GIF generation, including follow-up edits to a roller-skating monkey, and times a standard image at roughly 20 seconds. He then looks at Pruna AI’s open-weight PH3 image model, adapted from the MiniMax H3 video model. He covers file sizes of around 55 GB, VRAM needs from about 21 GB at 4-bit to 56 GB at 16-bit, which hardware can run it, and how it performs in a free Hugging Face Space.
The video also covers an open-weight video model that generates clips in about 30 seconds on Hugging Face and needs H100-class hardware for its quoted speeds. It also tests Meta’s Muse assistant, which now supports voice chats and video calls with customizable voices and connections to Gmail and calendar.
Viewers get a quick, practical picture of what’s new across image, video, and voice AI, plus a sense of which open models can realistically run on local hardware.
📺 Source: Matt Wolfe · Published November 06, 2026
🏷️ Format: Roundup
