---
license: other
license_name: minimax-h3-community-license-agreement
license_link: LICENSE
base_model: MiniMaxAI/MiniMax-H3
base_model_relation: adapter
pipeline_tag: image-text-to-video
library_name: minimax-h3
tags:
  - hyperflow-h3
  - lora
  - minimax-h3
  - video-generation
  - audio-video-generation
  - text-to-video
  - image-to-video
  - image-text-to-video
  - text-to-audio-video
  - image-to-audio-video
  - reference-to-audio-video
  - flow-matching
  - distillation
  - self-distillation
  - few-step
  - comfyui
  - custom-node
  - int8
---

# HyperFlow — ComfyUI node weights

<video
src="https://huggingface.co/drbaph/Hyperflow-Comfyui/resolve/main/hyperflow_1.mp4"
controls
loop
width="100%"> </video>

<video
src="https://huggingface.co/drbaph/Hyperflow-Comfyui/resolve/main/hyperflow_2.mp4"
controls
loop
width="100%"> </video>

<video
src="https://huggingface.co/drbaph/Hyperflow-Comfyui/resolve/main/hyperflow_3.mp4"
controls
loop
width="100%"> </video>

ComfyUI-converted builds of [HyperFlow](https://github.com/Video-Rebirth/hyperflow) (Video Rebirth's 8-step LoRA for MiniMax-H3, video + audio), for the **[ComfyUI-Hyperflow](https://github.com/Saganaki22/ComfyUI-Hyperflow)** custom node.

> **These are not the standalone LoRAs.** The files here require the custom node (they carry the time-embedder/endpoint LoRA + fixed 8-step grid the node consumes). If you want to use HyperFlow **without** the custom node, get the extracted standalone builds from [drbaph/MiniMax-H3-Turbo-Lora-ComfyUI](https://huggingface.co/drbaph/MiniMax-H3-Turbo-Lora-ComfyUI/) — those load with the stock `Load LoRA` node (backbone only, no two-time conditioning, output deviates from the released model).

![Screenshot 2026-09-19 081642](https://cdn-uploads.huggingface.co/production/uploads/63473b59e5c0717e6737b872/WoJrKW0CeXgcz6uq9Odv6.png)

## Files — download ONE into `models/hyperflow/`

| File | Size | Base | Notes |
| --- | --- | --- | --- |
| `custom_node_hyperflow_8step_v1.0_comfyui.safetensors` | 3.67 GiB | full (non-pruned) | the released 8-step model (two-time conditioning) |
| `custom_node_hyperflow_8step_v1.0_comfyui_pruned.safetensors` | 3.64 GiB | pruned/curve | backbone only, single-time (off-recipe) |

```text
📂 ComfyUI/
└── 📂 models/
    └── 📂 hyperflow/
        └── custom_node_hyperflow_8step_v1.0_comfyui.safetensors
```

- The `hyperflow.json` / `hyperflow_pruned.json` manifests ship with the node pack — you only download the `.safetensors`.
- Or skip downloading: enable the node's `download_if_missing` toggle and it fetches itself.
- Converted from the official `minimax_h3_hyperflow_8step_v1.0.safetensors` (SHA-256 `9297f450…df447`, verified); the original diffusers-layout file is rejected by the node, never translated at load time.

**Usage:** `Load Diffusion Model (MiniMax-H3) → ApplyHyperFlow → SamplerCustomAdvanced + euler` (feed the node's SIGMAS output). Optional sparse attention: core `Model Sparse Attention` (sol-attn: start_percent 0.25, dense_blocks "0,1", tau 1.0).

Weights are a Model Derivative of MiniMax-H3 under the [MiniMax H3 Community License](https://huggingface.co/videorebirth/hyperflow).