LynnReal-Omni beta 0.1: One Open Model for Video Generation and Editing

The LynnReal-AI team released beta 0.1 of the unified video model LynnReal-Omni, built on a 32B diffusion transformer with the MiniMax H3 architecture. It covers t2v, i2v, pose control, omni-reference, and video restoration, while the 27B Flash version generates 22 frames at 540p in 377 ms on a single H100.

Compiled by Sergey KostenchukPublished 2026-09-17Updated 2026-09-18

2026-09-17 Multimodal HuggingFace

🎨 LynnReal-Omni beta 0.1: One model for video generation and editing

The LynnReal-AI team released beta 0.1 of a unified video model built on a 32B diffusion transformer with the MiniMax H3 architecture. A single model covers t2v, i2v, pose control, omni-reference, style transfer, editing, and video restoration. The 27B Flash version generates 22 frames at 540p in 377 ms on an H100, and a 5-second 1344×768 clip with audio in 8.4 s.

🌍 Instead of a dozen specialized models, a single checkpoint accepts appearance references, 3D renders, and game captures: a step toward agentic visual workflows and real-time video streaming.

👤 Weights are on Hugging Face, with a ComfyUI port featuring the ComfyUI-LynnReal node pack and workflows. Testing was run on an H100 80GB and an RTX 4090 48GB (24 GB will not be enough); the Flash path does not work with clips longer than 11 seconds.

Source 1: https://huggingface.co/stdstu123/LynnReal-Onmi-beta-0.1 Source 2: https://github.com/LynnReal-AI/LynnReal-Omni

Sources