ComfyUI-Open-H3-Regenerate
Copyright 2026 rdh9616

This project includes fast_vae_decoder.py, copied unchanged from NVlabs/Sana
(https://github.com/NVlabs/Sana, branch sol-engine,
models/Open-H3-Regenerate-2K/open_h3/fast_vae/decoder.py), licensed under the
Apache License, Version 2.0. Copyright NVIDIA Corporation.

Model weights are not included. They keep their own licenses: the Open-H3-Regenerate-2K
refiner LoRA and fast H3 VAE decoder (MiniMax H3 Community License), LightX2V FL2VA Turbo,
LMS Sharp (Alissonerdx), and the H3 latent upscaler (LBH-123-AI).

TS adaptation (2026): optional native ComfyUI operations factory for INT8 convrot block linears.
