---
title: Open-H3-Regenerate-2K
canonical_url: "https://www.modelscope.cn/models/Efficient-Large-Model/Open-H3-Regenerate-2K"
md_url: "https://www.modelscope.cn/models/Efficient-Large-Model/Open-H3-Regenerate-2K.md"
repository: Efficient-Large-Model/Open-H3-Regenerate-2K
last_updated: 2026-10-28
license: other
base_model:
  - MiniMaxAI/MiniMax-H3
base_model_relation: adapter
parameters: 74.6B
tensor_type:
  - BF16
  - F32
library_name:
  - lora
  - safetensors
  - diffusers
  - pytorch
frameworks:
  - pytorch
downloads: 551
stars: 1
tags:
  - minimax-h3
  - lora
  - video-refinement
  - sol-engine
---

# Open-H3-Regenerate-2K

> Open-H3-Regenerate-2K - Efficient-Large-Model 在 ModelScope 开源的模型。Open-H3-Regenerate-2K

Efficient-Large-Model/Open-H3-Regenerate-2K 是 ModelScope 魔搭社区上的 74.6B 参数机器学习模型，采用 other 许可，基于 MiniMaxAI/MiniMax-H3 构建。

- **Repository**: Efficient-Large-Model/Open-H3-Regenerate-2K
- **License**: other
- **Parameters**: 74.6B
- **Base model**: MiniMaxAI/MiniMax-H3
- **Tags**: minimax-h3, lora, video-refinement, sol-engine
- **Downloads**: 551
- **Stars**: 1
- **Last updated**: 2026-10-28

Source: https://www.modelscope.cn/models/Efficient-Large-Model/Open-H3-Regenerate-2K

---

# Open-H3-Regenerate-2K

An open replacement for the closed H3-Regenerate-2K API: a three-step MiniMax-H3 refinement pass that regenerates a
768P H3 draft (1344×768, 121 frames) at **2560×1440**.

[Project page](https://nvlabs.github.io/Sana/Open-H3-Regenerate-2K/) ·
[Code](https://github.com/NVlabs/Sana/tree/sol-engine/models/Open-H3-Regenerate-2K)

## Speed on one B200

One 5 s clip, 768P latent → 2560×1440 video: latent upscale + 3-step refine + VAE decode, warm, median of 3 runs:

| Forward | VAE decoder | Total | Speedup |
|---|---|---:|---:|
| `standard_fwd` (dense attention, BF16) | released H3 VAE | 87.26 s | 1.00× |
| `sol_engine_fwd` (Sol-Attn, MXFP8, CUDA graphs) | released H3 VAE | 37.39 s | 2.33× |
| `sol_engine_fwd` (Sol-Attn, MXFP8, CUDA graphs) | fast H3 VAE decoder | 22.45 s | 3.89× |

The fast H3 VAE decoder is the non-causal 24-layer decoder of
[StreamRender-H3](https://nvlabs.github.io/Sana/Sol-Engine/StreamRender-H3/), distilled from the released H3 VAE;
it decodes the 2560×1440 clip in 1.6 s instead of 15 s. Latent upscale + refine alone go from 73.11 s to 20.46 s (3.57×).

Per-module numbers are in the code README.

## Files

| Path | Content |
|---|---|
| `lora/open_h3_regenerate_2k_lora.safetensors` | One merged LoRA (FP32, rank 576 on 312 linears, scale 1) |
| `lora/sharp_curve_branches.safetensors` | The 50 LMS Sharp curve AdaLN branches (FP32) |
| `lora/lora_config.json`, `lora/module_ranks.json` | Sources, hashes, per-module ranks |
| `lora/MERGE_VERIFY.json` | Merge check against the three separate adapters |
| `latent_upsampler/` | Author 3D-conv latent upscaler, BF16 |
| `vae_fast/fast_h3_vae_noncausal_24l.safetensors` | Fast H3 VAE decoder from StreamRender-H3 (non-causal, 24 layers, 1.62B parameters) |
| `base_model/` | MiniMax-H3 Diffusers components (transformer, VAE, audio VAE, text encoder, tokenizer, processor, schedulers), unmodified copy of revision `6818f6c` |

`base_model/` mirrors [MiniMaxAI/MiniMax-H3 @ 6818f6c](https://huggingface.co/MiniMaxAI/MiniMax-H3/tree/6818f6c32d12b210915e44ad56a4228c2608f160)
without changes (the native `FL2VA/`, `Ref2VA/` and `transformer_ref/` folders are left out). `download.py` in the code
folder fetches everything here plus the optional HyperFlow 8-step adapter and Sharp's 1025×8 `adaln_t_table`
(a base-derived tensor, read by byte range from `Comfy-Org/MiniMax-H3`).

## What the LoRA contains

The reference recipe applies three adapters to the MiniMax-H3 transformer:

| Adapter | Rank / alpha |
|---|---|
| Open-H3 refiner, step 5000 EMA | 384 / 384 |
| LightX2V FL2VA Turbo v1.2 | 128 / 8 |
| LMS Sharp v1.0 | 64 |

They are merged by concatenating the factors along the rank axis with the strengths (and Turbo's alpha/rank) folded
into `B`, so `B @ A` equals the sum of the three scaled deltas exactly. Over all 312 linears the merged delta
matches the separate adapters to a relative error of 2.2e-8 (`MERGE_VERIFY.json`). Sharp's curve branches take an
8-D timestep curve instead of the layer input, so they are kept as separate modules.

Sampling: three Euler steps on sigmas `[0.909375, 0.857143, 0.666667, 0]`, CFG 1, initial state
`(1 - 0.909375) * upscaled + 0.909375 * noise`, model time `1 - sigma`.

## License

The refiner LoRA, the fast H3 VAE decoder and `base_model/` follow the
[MiniMax H3 Community License](https://huggingface.co/MiniMaxAI/MiniMax-H3/blob/6818f6c32d12b210915e44ad56a4228c2608f160/LICENSE),
including its territory, use and redistribution terms. The components we did not train follow their own licenses:
the [latent upscaler](https://github.com/LBH-123-AI/Comfyui_Minimax_h3_latent_Upscaler), and the
[LightX2V FL2VA Turbo](https://huggingface.co/lightx2v/Minimax-h3-Turbo/tree/3ec17a324ced54151364f24f8b5fb6bf7e26414f) and
[LMS Sharp](https://huggingface.co/Alissonerdx/Minimax-H3-ComfyUI/blob/main/loras/minimax_h3_lms_v1.0_r64.safetensors)
LoRAs merged into our LoRA. This bundle is not
covered by a single blanket license. The runtime code's license does not replace
model licenses.