Repository navigation
Add Open-H3-Regenerate-2K-IC: in-context regeneration of an H3 draft at 1080P / 2K / 4K - #584
Merged
HaozheLiu-ST merged 3 commits intoNov 9, 2026
Conversation
…at 1080P / 2K / 4K (inference) models/Open-H3-Regenerate-2K-IC/: inference code for the IC version of Open-H3-Regenerate-2K. The low-resolution draft stays in the context of MiniMax-H3's Ref2VA transformer (transformer_ref) as a reference video and an IC LoRA (step-6000 EMA, rank 384) generates the high-resolution clip, keeping faces, lip sync and lettering of the draft. - infer.py: torchrun on 1/2/4/7/8 GPUs (FSDP + Ulysses context parallelism); samplers up3 (default: latent upscaler, noised to sigma 0.835, last 3 HyperFlow 8-step points), up4 and hf8; LMS Sharp 1.0; 1080P / 2K / 4K and 384P -> 768P; fast H3 VAE decoder v2 (FlashAttention-4 engine on Blackwell, PyTorch elsewhere) or the released H3 VAE. - download.py: Efficient-Large-Model/Open-H3-Regenerate-2K-IC plus MiniMax-H3 @ 42ed227 from the official repository. - merge_lora.py: IC LoRA + LMS Sharp as one LoRA for runtimes that load one LoRA per model. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…te-2K/IC The IC inference code now lives in the existing Open-H3-Regenerate-2K folder (IC/); its README links are relative to that folder, and the Open-H3-Regenerate-2K README points to it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…itory (unmodified MiniMax-H3 @ 42ed227) download.py fetches base_model/ from Efficient-Large-Model/Open-H3-Regenerate-2K-IC by default, as Open-H3-Regenerate-2K does; --base-from-minimax takes it from the official MiniMaxAI/MiniMax-H3 at the same revision instead. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
HaozheLiu-ST
self-requested a review
November 9, 2026 05:03
HaozheLiu-ST
approved these changes
Nov 9, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds
models/Open-H3-Regenerate-2K/IC/(inference code only) inside the existing Open-H3-Regenerate-2K folder; the Open-H3-Regenerate-2K README links to it.The low-resolution MiniMax-H3 draft (768P) stays in the context of the Ref2VA transformer (
transformer_ref) as a reference video, and an in-context LoRA (step-6000 EMA, rank 384) generates the high-resolution version of the same clip, keeping the draft's faces, lip sync and lettering.infer.py:torchrunon 1, 2, 4, 7 or 8 GPUs (FSDP + Ulysses context parallelism). Samplersup3(default: latent upscaler, noised to σ 0.835, last 3 points of the HyperFlow 8-step grid),up4andhf8; LMS Sharp 1.0;--resolution 1080p / 2k / 4k, and768pfrom a 384P / 512P / 576P draft; fast H3 VAE decoder v2 (FlashAttention-4 engine on Blackwell, plain PyTorch elsewhere) or the released H3 VAE.download.py: the IC weights from Efficient-Large-Model/Open-H3-Regenerate-2K-IC andbase_model/(MiniMax-H3 @42ed227components, unmodified copy in that repository;--base-from-minimaxtakes it from the official repository instead). Weights PR: https://huggingface.co/Efficient-Large-Model/Open-H3-Regenerate-2K-IC/discussions/1merge_lora.py: IC LoRA + LMS Sharp as one exact LoRA, for runtimes that load one LoRA per model.