MAGI-1 (Sand AI)
What they did
An autoregressive video model that generates video chunk by chunk, which suits streaming and extension. Weights and inference code are open, including 4.5B distilled and quantised variants and ComfyUI nodes.
Worth knowing
- Price: Open weights on Hugging Face (apache-2.0)
- License: apache-2.0
- Status: sourced. We have not run this ourselves yet (no "Tested" badge).
- Quality bar (computed by code): 4/5 (primary source ✓, reproducible ✗, new ✓, example output ✓, safe ✓)
Get Atlas Dispatch
The best new AI video workflows and prompts, weekly.
Weekly: 5 new prompts and workflows. Unsubscribe anytime.