aryan5v · Oct 29, 2026 · Oct 29, 2026 · Oct 29, 2026 · Oct 29, 2026 · Oct 30, 2026
diff --git a/docs-site/content/cookbook/minimax-h3/open-h3-regenerate-2k.mdx b/docs-site/content/cookbook/minimax-h3/open-h3-regenerate-2k.mdx
@@ -0,0 +1,7 @@
+---
+title: Open-H3-Regenerate-2K
+description: Configure Open-H3-Regenerate-2K serving, which regenerates a 768P MiniMax-H3 draft at 2560x1440.
+mode: frame
+---
+
+<Cookbook modelId="open-h3-regenerate-2k" />
diff --git a/docs-site/lib/cookbook-pages.ts b/docs-site/lib/cookbook-pages.ts
@@ -48,6 +48,15 @@ export const recipePages = [
     tags: ["Reference to video", "Audio", "NVFP4"],
     featured: true,
   },
+  {
+    id: "open-h3-regenerate-2k",
+    family: "MiniMax H3",
+    title: "Open-H3-Regenerate-2K",
+    mark: "H3",
+    path: "/cookbook/minimax-h3/open-h3-regenerate-2k",
+    description: "Regenerate a 768P MiniMax-H3 draft at 2560x1440 with a three-step refiner.",
+    tags: ["Video to video", "2K", "3-step"],
+  },
   {
     id: "fastwan21",
     family: "Wan",
@@ -124,6 +133,7 @@ const backendModels: Record<string, string> = {
   "FastVideo/FastVideo-Minimax-FastH3-Preview-v0.2": "fasth3-preview",
   "FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree": "fasth3-4step",
   "FastVideo/FastH3-OmniRef-NVFP4-QAD": "fasth3-omniref",
+  "Efficient-Large-Model/Open-H3-Regenerate-2K": "open-h3-regenerate-2k",
   "FastVideo/FastWan2.1-T2V-1.3B-Diffusers": "fastwan21",
   "Wan-AI/Wan2.1-I2V-14B-480P-Diffusers": "wan21-i2v",
   "FastVideo/LTX2-Distilled-Diffusers": "ltx2-distilled",

diff --git a/docs-site/navigation.json b/docs-site/navigation.json
@@ -174,6 +174,10 @@
             "label": "FastH3 4-Step V1",
             "href": "/cookbook/minimax-h3/fasth3-4step"
           },
+          {
+            "label": "Open-H3-Regenerate-2K",
+            "href": "/cookbook/minimax-h3/open-h3-regenerate-2k"
+          },
           {
             "label": "Recipes and setup",
             "href": "/cookbook/minimax-h3"

diff --git a/docs/assets/cookbook-recipes.json b/docs/assets/cookbook-recipes.json
@@ -760,6 +760,22 @@
       ],
       "limitations": ["Pass --reference-audio for an additional audio reference. The checked-in source defaults to four GPUs."]
     },
+    {
+      "id": "open-h3-regenerate-2k",
+      "family": "minimax_h3",
+      "stage": "inference",
+      "task": "Video super-resolution (2K regeneration)",
+      "label": "Open-H3-Regenerate-2K",
+      "summary": "Regenerate a 768P MiniMax-H3 draft at 2560x1440 with NVlabs' three-step refiner. The output is silent.",
+      "model": "Efficient-Large-Model/Open-H3-Regenerate-2K",
+      "source": "examples/inference/basic/basic_open_h3_regenerate_2k.py",
+      "command": ": \"${INPUT_VIDEO:?Set INPUT_VIDEO to a 768P MiniMax-H3 draft}\"\npython examples/inference/basic/basic_open_h3_regenerate_2k.py --video \"$INPUT_VIDEO\" --prompt \"A red fox trots through fresh snow in a pine forest at golden hour.\"",
+      "gpu_types": ["NVIDIA"],
+      "hardware": {"platform": "cuda", "gpu_count": 1, "evidence": "source-configured"},
+      "evidence": "Source-backed",
+      "expected_artifact": "2560x1440 MP4 at outputs/open_h3_regenerate_2k/regen2k_0.mp4",
+      "related": ["minimax-h3-t2v", "fasth3-8step-v2-cuda"]
+    },
     {
       "id": "fasth3-lora-preview",
       "family": "minimax_h3",

diff --git a/docs/cookbook/options.yaml b/docs/cookbook/options.yaml
@@ -44,7 +44,7 @@ runtimes:
         host: 0.0.0.0
         port: 8000
         served_model_name: null
-    workloads: [t2v, i2v]
+    workloads: [t2v, i2v, v2v]
     options: &cuda_options
       server.served_model_name:
         anyOf:

diff --git a/docs/cookbook/recipes/open-h3-regenerate-2k.yaml b/docs/cookbook/recipes/open-h3-regenerate-2k.yaml
@@ -0,0 +1,18 @@
+title: Open-H3-Regenerate-2K
+summary: Regenerate a 768P MiniMax-H3 draft at 2560x1440 with three refine steps.
+
+deployments:
+  cuda-rest:
+    runtime: fastvideo-cuda-rest
+    guide: docs/inference/open_h3_regenerate_2k.md
+    workload: v2v
+    defaults: examples/serving/openai_open_h3_regenerate_2k.yaml
+    requirements:
+      - 'UV_TORCH_BACKEND=cu130 uv pip install -e ".[fasth3]"'
+      - "One GPU with about 150 GB free (B200 or GB200). Send the draft as video_reference; local paths must sit under server.media_roots."
+      - "Optional NVFP4 decoder (Blackwell only): hf download FastVideo/FastH3-NVFP4-VAE-Decoder --include 'quality-36L/decoder-nvfp4.safetensors' --local-dir fasth3-nvfp4-vae"
+    options:
+      - server.*
+      - generator.engine.offload.text_encoder
+      - generator.pipeline.experimental.h3_vae_nvfp4_decoder
+      - default_request.sampling.seed
diff --git a/docs/design/inference_schema_parity_inventory.yaml b/docs/design/inference_schema_parity_inventory.yaml
@@ -153,6 +153,12 @@ surfaces:
 
   pipeline_config_extensions:
     preset_owned:
+      draft_size:
+        sources:
+          - fastvideo.configs.pipelines.open_h3_regenerate_2k.OpenH3Regenerate2KPipelineConfig
+      refine_start_sigma:
+        sources:
+          - fastvideo.configs.pipelines.open_h3_regenerate_2k.OpenH3Regenerate2KPipelineConfig
       audio_encoder_precision:
         sources:
           - fastvideo.configs.pipelines.wan.WanS2V14BConfig

diff --git a/docs/inference/open_h3_regenerate_2k.md b/docs/inference/open_h3_regenerate_2k.md
@@ -0,0 +1,135 @@
+# Open-H3-Regenerate-2K
+
+`OpenH3Regenerate2KPipeline` regenerates a 768P [MiniMax-H3](https://huggingface.co/MiniMaxAI/MiniMax-H3) draft video
+(1344x768, 121 frames, 24 fps) at **2560x1440** with three refine steps. It is video to video: the draft is the input,
+and the output is a silent 2K MP4. FastVideo ports NVlabs'
+[Open-H3-Regenerate-2K](https://huggingface.co/Efficient-Large-Model/Open-H3-Regenerate-2K)
+([project page](https://nvlabs.github.io/Sana/Open-H3-Regenerate-2K/),
+[reference code](https://github.com/NVlabs/Sana/tree/sol-engine/models/Open-H3-Regenerate-2K)); the dense BF16
+`standard_fwd` path of that repo is the parity reference. The accelerated `sol_engine_fwd` path (Sol-Attn, MXFP8, CUDA
+graphs) and the fast 24-layer VAE decoder are not ported.
+
+```text
+draft video -> H3 VAE encode -> latent upscaler (2x, 3D conv) -> 3 Euler refine steps -> H3 VAE decode -> 2560x1440
+                                                 H3 DiT + merged LoRA + Sharp curve AdaLN branches
+```
+
+## What runs
+
+| Step | Details |
+|---|---|
+| Draft encode | `ffmpeg` resamples to 24 fps, cover-scales (bilinear) and centre-crops to 1344x768 (`draft_size: 768p`, or 1024x576 for `576p`), holds the last frame to 121 frames plus 3 copies. The FP32 H3 VAE posterior is sampled under seed 42. |
+| Latent upscale | The author's 3D-conv upscaler (`fastvideo/models/upsamplers/minimax_h3_latent_upscaler.py`), BF16, to the 160x90 latent grid. |
+| Refine | Euler on sigmas `[0.909375, 0.857143, 0.666667, 0]`, CFG 1, model time `1 - sigma`, initial state `(1 - 0.909375) * upscaled + 0.909375 * noise`. `num_inference_steps: 2` uses `[0.7, 0.494118, 0]` (softer, about 30% faster). |
+| Transformer | The base MiniMax-H3 DiT (the same module as the other H3 pipelines) with the released merged LoRA (FP32, rank 576 on 312 linears, scale 1) and 50 LMS Sharp curve AdaLN branches. |
+| Decode | The released H3 VAE; the output is trimmed to 121 frames. |
+
+The merged LoRA loads through the standard H3 community-LoRA path as a constructor-time adapter: the pipeline sets
+`lora_path` to `<model>/lora/open_h3_regenerate_2k_lora.safetensors` when you do not pass one. The Sharp branches take an
+8-D timestep curve (`lora/adaln_t_table.bin`, 1025x8) instead of the layer input, so they cannot be folded into weights;
+`sharp_curve.py` adds them to the AdaLN projection outputs with forward hooks.
+
+**Audio.** The reference feeds an encoded-silence audio stream to the DiT as conditioning and writes the video with no
+audio track (`-an`); it neither regenerates nor passes through the draft's audio. FastVideo does the same: the draft's
+soundtrack is ignored and the output MP4 has no audio.
+
+## Weights
+
+The release repo keeps the unmodified MiniMax-H3 components (a copy of `MiniMaxAI/MiniMax-H3` at `6818f6c`) under
+`base_model/` next to `lora/` and `latent_upsampler/`. Point `model_path` at the Hub id or at a local copy of that layout.
+`adaln_t_table.bin` is base-derived and is not redistributed in the release repo: put the 32,800-byte file (SHA-256
+`c02a6c11888297688c1e6278185ea1f947023acfc69f9003bbcdcec9a229a8e7`) at `lora/adaln_t_table.bin`. The reference
+`download.py` fetches it by byte range from `Comfy-Org/MiniMax-H3` at revision
+`e5eb578a89295337b8ff433a035929ce0279e0b6` (tensor `adaln_t_table` of
+`diffusion_models/minimax_h3_ref2va_pruned_int8_convrot.safetensors`) and verifies the hash.
+
+A GB200/B200-class GPU is needed (about 150 GB with the DiT, the 2K activations and the VAE).
+
+## Usage
+
+```bash
+python examples/inference/basic/basic_open_h3_regenerate_2k.py \
+    --model-path Efficient-Large-Model/Open-H3-Regenerate-2K \
+    --video draft_768p.mp4 --prompt "A red fox trots through fresh snow" --output outputs/regen2k
+```
+
+The prompt conditions the refine steps (use the draft's own prompt). Drafts from other video models are better encoded at
+576P (`--draft-size 576p`). `--steps 2` runs the two-step variant. The output is always 2560x1440; the first 121 frames of
+the draft are used.
+
+## Serving
+
+```bash
+fastvideo serve --config examples/serving/openai_open_h3_regenerate_2k.yaml
+```
+
+`examples/serving/openai_open_h3_regenerate_2k.yaml` loads the release on one GPU, sets `media_roots`, and pins the 2K
+defaults. The draft arrives as the typed `video_reference` of `POST /v1/videos/sync` (the response body is the MP4) or of
+`POST /v1/videos` (an async job; poll `GET /v1/videos/{id}` and download `GET /v1/videos/{id}/content`):
+
+```bash
+curl -sS -X POST http://localhost:8000/v1/videos/sync \
+  -H 'Content-Type: application/json' \
+  -d '{
+        "prompt": "A red fox trots through fresh snow",
+        "video_reference": {"video_url": "https://example.com/draft_768p.mp4"},
+        "seed": 1766580932
+      }' \
+  -o regen_2k.mp4 -D -
+```
+
+The response headers carry `X-Request-Id`, `X-Model`, `X-Inference-Time-S` and `X-Stage-Durations`. Rules:
+
+- Exactly one `video_reference` (or the legacy `video_path` / `video_url`) and no image or audio references.
+- `http(s)` URLs must resolve to public addresses unless `server.allow_private_media_urls` is set. A local path is accepted
+  only inside a directory listed in `server.media_roots`. `data:` URLs are refused for video references; this policy is
+  unchanged.
+- `size`, `width` and `height` may be omitted or `2560x1440`; `num_frames` may be omitted or 124 (the frame count of the
+  VAE grid; the video has 121 frames); `fps` must be 24. `seconds` is not accepted.
+- `num_inference_steps` is 3 (default) or 2; `seed` sets the refine noise.
+- Serve with `attention_backend: TORCH_SDPA`: the refine stage selects cuDNN attention, which at about 133k tokens is
+  about 3x faster than PyTorch's default flash dispatch. The `FLASH_ATTN` backend bypasses that selection.
+
+## Parity with the reference
+
+Measured on one GB200 (BF16, `TORCH_SDPA` with cuDNN attention) against the reference `standard_fwd` with the released
+VAE, same draft, seed (1766580932), prompt and VAE posterior seed (42). The reference's noise is a CUDA generator draw
+that FastVideo reproduces (the same `torch.Generator(device="cuda")` seed), so no injected noise was needed.
+
+| Stage / quantity | Fox draft (FastVideo H3 t2v, 768P) | Dance draft (576P, project page) |
+|---|---|---|
+| Draft latent, upscaled latent, DiT input | bit-identical | bit-identical |
+| One DiT forward (LoRA + Sharp), relative L2 | 1.1e-2 | 2.6e-2 |
+| Refined latent, relative L2 | 1.3e-2 | 3.7e-2 |
+| Decoded video PSNR / SSIM (21 frames) | 45.2 dB / 0.993 | 42.1 dB / 0.992 |
+
+The differences come from text-encoder rounding (relative L2 1e-3 on the prompt embeddings) and BF16 kernel choices;
+the reference reports the same order (1.4% relative L2 on the refined latent between its own runs).
+
+Warm timings, one GB200 (seconds):
+
+| | Upscale | Refine (3 steps) | VAE decode | Total |
+|---|---:|---:|---:|---:|
+| Reference `standard_fwd` | 0.87 | 71.6 | 15.0 | 87.5 |
+| FastVideo (end to end, incl. draft encode 5-10 s and decode to host) | 5.9-10.7 (encode + upscale + silence) | 69.5 | 25.0 | 107.5 |
+
+The first request builds nothing extra but loads weights (about 70 s once). The decode stage is slower than the
+reference because it also copies 121 frames to host memory and the MP4 encode is outside both columns.
+
+## Licences
+
+The release is not covered by one licence; its `README.md` says so, and the runtime code's licence does not replace the
+model licences. FastVideo only loads these weights and does not redistribute them.
+
+- The refiner LoRA, the fast H3 VAE decoder and `base_model/` (a copy of `MiniMaxAI/MiniMax-H3` at `6818f6c`: DiT, VAEs,
+  text encoder) follow the
+  [MiniMax H3 Community License](https://huggingface.co/MiniMaxAI/MiniMax-H3/blob/6818f6c32d12b210915e44ad56a4228c2608f160/LICENSE),
+  including its territory, use and redistribution terms (`base_model/LICENSE`, `NOTICE`). The Qwen3-VL text encoder's
+  Apache-2.0 text ships as `LICENSE-Qwen.txt`.
+- The latent upscaler follows its own licence:
+  [LBH-123-AI/Comfyui_Minimax_h3_latent_Upscaler](https://github.com/LBH-123-AI/Comfyui_Minimax_h3_latent_Upscaler)
+  (MIT, per the reference code) and its [weights](https://huggingface.co/LBH-123-AI/Minimax_h3_latent_Upscaler)
+  (`latent_upsampler/LICENSE`, Apache-2.0). `minimax_h3_latent_upscaler.py` is a port of that node's network.
+- The merged LoRA includes [LightX2V FL2VA Turbo](https://huggingface.co/lightx2v/Minimax-h3-Turbo) and
+  [LMS Sharp](https://huggingface.co/Alissonerdx/Minimax-H3-ComfyUI), which keep their own licences; the Sharp
+  `adaln_t_table` is read from [Comfy-Org/MiniMax-H3](https://huggingface.co/Comfy-Org/MiniMax-H3).
diff --git a/docs/inference/support_matrix.md b/docs/inference/support_matrix.md
@@ -68,6 +68,7 @@ column links a runnable script in `examples/inference/basic/` where one exists.
 | matrixgame | `FastVideo/Matrix-Game-2.0-Base-Distilled-Diffusers`<br />`FastVideo/Matrix-Game-2.0-GTA-Distilled-Diffusers`<br />`FastVideo/Matrix-Game-2.0-TempleRun-Distilled-Diffusers`<br />`FastVideo/Matrix-Game-2.0-Base-Diffusers`<br />`FastVideo/Matrix-Game-2.0-GTA-Diffusers`<br />`FastVideo/Matrix-Game-2.0-TempleRun-Diffusers`<br />`mignonjia/mg_longtuning_distilled_zelda`<br />`mignonjia/mg_sf_distilled_zelda_1k_steps`<br />`mignonjia/mg_sf_distilled_zelda`<br />`mignonjia/mg_causal_zelda`<br />`mignonjia/mg_bidirectional_zelda` | I2V | [basic_matrixgame2.py](https://github.com/hao-ai-lab/FastVideo/blob/main/examples/inference/basic/basic_matrixgame2.py) |
 | matrixgame | `FastVideo/Matrix-Game-3.0-Base-Distilled-Diffusers` | I2V | [basic_matrixgame3.py](https://github.com/hao-ai-lab/FastVideo/blob/main/examples/inference/basic/basic_matrixgame3.py) |
 | minimax_h3 | `MiniMaxAI/MiniMax-H3` | T2V, I2V | [T2VA](https://github.com/hao-ai-lab/FastVideo/blob/main/examples/inference/basic/basic_minimax_h3_t2v.py)<br />[FL2VA](https://github.com/hao-ai-lab/FastVideo/blob/main/examples/inference/basic/basic_minimax_h3_fl2va.py)<br />[Ref2VA](https://github.com/hao-ai-lab/FastVideo/blob/main/examples/inference/basic/basic_minimax_h3_ref2va.py) |
+| open_h3_regenerate_2k | `Efficient-Large-Model/Open-H3-Regenerate-2K` | — | [basic_open_h3_regenerate_2k.py](https://github.com/hao-ai-lab/FastVideo/blob/main/examples/inference/basic/basic_open_h3_regenerate_2k.py) |
 | sd35 | `stabilityai/stable-diffusion-3.5-medium` | T2I | [basic_sd35_t2i.py](https://github.com/hao-ai-lab/FastVideo/blob/main/examples/inference/basic/basic_sd35_t2i.py) |
 | qwen_image21 | `Qwen/Qwen-Image-2.1` | T2I, I2I | [basic_qwen_image21.py](https://github.com/hao-ai-lab/FastVideo/blob/main/examples/inference/basic/basic_qwen_image21.py) |
 | stable_audio | `FastVideo/stable-audio-open-1.0-Diffusers` | T2V | [basic_stable_audio.py](https://github.com/hao-ai-lab/FastVideo/blob/main/examples/inference/basic/basic_stable_audio.py) |
@@ -121,6 +122,9 @@ and distilled) generate video with audio from text, optionally plus an image (se
 `kandinsky6_sr` IDs upscale an existing video (see the
 [Video SR guide](kandinsky6_sr.md)).
 
+**Note (Open-H3-Regenerate-2K)**: `open_h3_regenerate_2k` regenerates an existing 768P MiniMax-H3 draft video at
+2560x1440 (video to video, silent output); see the [Open-H3-Regenerate-2K guide](open_h3_regenerate_2k.md).
+
 **Note (LTX-2.5)**: T2V and I2V produce synchronized video and audio. The
 official `Lightricks/LTX-2.5` repository uses gated, split component files, so
 it must first be converted to FastVideo's component layout. See the

diff --git a/docs/js/cookbook-workloads.mjs b/docs/js/cookbook-workloads.mjs
@@ -32,6 +32,14 @@ export const WORKLOADS = Object.freeze({
     extension: "wav",
     media_type: "audio/wav",
   }),
+  v2v: Object.freeze({
+    input: "video",
+    output: "video",
+    endpoint: "/v1/videos/sync",
+    extension: "mp4",
+    media_type: "video/mp4",
+    reference: "video_reference",
+  }),
   v2a: Object.freeze({
     input: "video",
     output: "audio",

diff --git a/docs/tests/cookbook-config.test.mjs b/docs/tests/cookbook-config.test.mjs
@@ -488,6 +488,7 @@ test("sample requests follow shared workload media", () => {
     t2i: { endpoint: /\/v1\/images\/generations/, output: /--output output\.json/, reference: null },
     t2a: { endpoint: /\/v1\/videos\/sync/, output: /--output output\.wav/, reference: null },
     v2a: { endpoint: /\/v1\/videos\/sync/, output: /--output output\.wav/, reference: "video_reference" },
+    v2v: { endpoint: /\/v1\/videos\/sync/, output: /--output output\.mp4/, reference: "video_reference" },
   };
   for (const [workload, expected] of Object.entries(expectations)) {
     const data = catalog();