---
license: other
license_name: minimax-h3-community
license_link: LICENSE
base_model: MiniMaxAI/MiniMax-H3
library_name: diffusers
pipeline_tag: text-to-video
tags:
- text-to-video
- video
- audio
- text-to-audio-video
- distillation
- dmd2
- few-step
- minimax-h3
- fastvideo
- fasth3
---

# FastVideo-FastH3-4-step-v1

Private FastVideo evaluation snapshot of the v10.5 data-free DMD2 MiniMax-H3
student at training step 1300. It generates synchronized video and audio with
four transformer forwards.

This repository is intentionally private. It is an internal checkpoint for
evaluation and UI integration, not a public release recommendation.

## Checkpoint identity

- Training run: `20260823T132231Z-v10p5-datafree-sp4`
- Step: 1300
- Transformer content SHA-256: `b36987515e4c75fa4c7aaa632a7842c829ea141b235358a54d782b51230497b3`
- Checkpoint metadata SHA-256: `dcad0fbee2a7c7e75e53435f4fd98fccf3138844883874edf057962ab48fa428`
- FastVideo execution commit: `48a047c05ff4138f20cfa33351499c6ec5945f5d`

`checkpoint_content.json` records every indexed transformer shard and digest.
`checkpoint_metadata.json` preserves the source training configuration.

## Contents

This is a self-contained Diffusers modular-pipeline package. The distilled
bf16 student is under `transformer/`; the text encoder, tokenizer, processor,
video VAE, audio VAE, and schedulers are unmodified MiniMax-H3 components.

The `transformer_ref` component is not bundled because this student targets
T2AV. Its modular-index entry remains on `MiniMaxAI/MiniMax-H3`.

## Exact FastVideo inference contract

The student was trained on the explicit ladder `[999, 749, 500, 250]`. These
are four transformer forwards; do not replace the ladder with an arbitrary
four-point scheduler grid.

```bash
export FASTVIDEO_DMD_DENOISING_STEPS=999,749,500,250

python examples/inference/basic/basic_fasth3.py \
  --model-path FastVideo/FastVideo-FastH3-4-step-v1 \
  --prompt '<MiniMax-H3 T2AV prompt>' \
  --steps 5 \
  --num-gpus 4 \
  --vsa-sparsity 0.9 \
  --vsa-tile-size 64 \
  --vsa-kernel sm100a
```

The five scheduler grid points execute the four trained jumps. Use guidance
scale 1.0, VSA-H3 at 90% sparsity with 64-token tiles, and SP=4/FSDP across one
four-GPU GB200 tray for the validated route. The exact v10.5 execution commit
runs the sparse DiT eagerly; regional compile is disabled for this snapshot.

## Training summary

- Method: data-free DMD2 with carried backward-simulation rollouts.
- Student grid: `[999, 749, 500, 250]`.
- Student attention: VSA-H3, 64-token tiles, 90% sparsity.
- Prompt conditioning: mixed native resolution and duration T2AV prompts; no
  target video latents are consumed by the data-free objective.
- Global batch size: 64.
- Student and critic learning rate: `2e-6`.
- Precision: fp32 training state and bf16 inference export.
- Training topology: 32 GB200 GPUs, SP=4, full-world FSDP sharding.

## Limitations

- Intermediate checkpoint at step 1300 of a planned 4000-step run.
- T2AV only; the reference-conditioned transformer was not distilled here.
- Quality remains below base MiniMax-H3 on difficult motion, fine structure,
  and some audio details.
- Sampling outside the trained ladder or attention geometry is
  off-distribution.
- Inherits the MiniMax H3 Community License and base-model limitations.