MiniMax H3 — light video VAE decoder (diffusers + ONNX)
Repository: corechan/MiniMax-H3-LightVAE

MiniMax H3 is licensed under the MiniMax H3 Community License Agreement,
Copyright © 2026 MiniMax. All Rights Reserved.

Source: the light video VAE decoder of LynnReal-Omni by LynnReal-AI
(https://huggingface.co/stdstu123/LynnReal-Onmi-beta-0.1 ,
comfyui/models/vae/lynnreal_omni_light_vae_fp16.safetensors ;
https://github.com/LynnReal-AI/LynnReal-Omni), a distilled derivative of the
MiniMax H3 video VAE decoder, distributed under the MiniMax H3 Community License
Agreement.

Modifications: converted from the ComfyUI key layout to the diffusers layout
(fused to_qkv split into to_q / to_k / to_v), stored in fp16, and one decoding
tile exported to ONNX (RMSNorm rewritten with basic operations). Not retrained
or fine-tuned.

This is an unofficial community upload. It is not affiliated with, endorsed by,
or supported by MiniMax or LynnReal-AI.