# minimax_h3_ref2va_pruned_fp8_scaled.safetensors

> MiniMax H3 Ref2VA pruned (fp8 scaled) · Reference-to-video diffusion model for MiniMax H3

This is MiniMax H3’s Ref2VA model, pruned and stored in fp8 scaled. It makes video with sound from your prompt plus reference images of characters, objects or places. Put it in ComfyUI/models/diffusion_models and load it with Load Diffusion Model.

Civitai names this download minimaxH3PrunedFp8_ref2vaPrunedFp8.safetensors. Same bytes. Comfy-Org suggests int8_convrot instead when your PyTorch supports it.

## Facts

- **Type:** Diffusion model (video + audio)
- **Format:** Safetensors · fp8 scaled
- **Size:** 20.96 GB (20,958,205,608 bytes)
- **SHA256:** f86f2f79ebd2d76eb8eeb46091e83982e6ff51d255747e7b16e92834b392b8e9
- **ComfyUI folder:** ComfyUI/models/diffusion_models/
- **Loader node:** Load Diffusion Model
- **License:** MiniMax H3 Community License

## What is minimax_h3_ref2va_pruned_fp8_scaled.safetensors?

MiniMax H3 ships as two diffusion models: FL2VA for text, image and first-last-frame to video, and Ref2VA for reference-guided video. This file is the Ref2VA one. You give it up to several reference images and a prompt, and it keeps those characters or objects consistent through the clip while generating matching audio.

“Pruned” means Comfy-Org removed weights the model does not need at inference, which drops the full model from 66 GB to 40 GB in bf16. “fp8 scaled” halves it again to 20.96 GB. Comfy-Org recommends this fp8 file only if you cannot use the int8_convrot version.

Civitai’s listing of the same file (model 2832970) renames it minimaxH3PrunedFp8_ref2vaPrunedFp8.safetensors on download. The Civitai API hash matches the Hugging Face file.

## Download and install

- [Hugging Face · Comfy-Org/MiniMax-H3](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_ref2va_pruned_fp8_scaled.safetensors) · saves as `minimax_h3_ref2va_pruned_fp8_scaled.safetensors` · ComfyUI repackage of the official weights. Recommended.
- [Civitai · MiniMax H3 pruned fp8 scaled](https://civitai.com/models/2832970) · saves as `minimaxH3PrunedFp8_ref2vaPrunedFp8.safetensors` · Same bytes under Civitai’s download name. May ask you to sign in.

```sh
cd ComfyUI/models/diffusion_models
wget -O minimax_h3_ref2va_pruned_fp8_scaled.safetensors \
  https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_ref2va_pruned_fp8_scaled.safetensors
sha256sum minimax_h3_ref2va_pruned_fp8_scaled.safetensors
# f86f2f79ebd2d76eb8eeb46091e83982e6ff51d255747e7b16e92834b392b8e9
```

Safetensors files hold tensors only and cannot run code. Still compare the SHA256: a failed download often leaves an HTML error page saved under the model’s filename.

## How to load it in ComfyUI

Load Diffusion Model reads the diffusion model and outputs a MODEL for the sampler. The text encoder and VAE load in their own nodes.


## Files a MiniMax H3 video workflow needs

Use the Ref2VA diffusion model instead of FL2VA for reference-to-video workflows.

- [`qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors`](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors) · Text encoder (Qwen3-VL 32B) · `models/text_encoders/` · 15.69 GB
- [`minimax_h3_video_vae_fp16.safetensors`](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors) · Video VAE · `models/vae/` · 5.21 GB
- [`minimax_h3_audio_vae_fp32.safetensors`](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors) · Audio VAE · `models/vae/` · 605 MB
- [`minimax_h3_fl2v_turbo_4step_v1.0_768p_comfyui_bf16.safetensors`](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/loras/minimax_h3_fl2v_turbo_4step_v1.0_768p_comfyui_bf16.safetensors) · Turbo LoRA, 4 steps · `models/loras/` · 1.96 GB

## Workflows: Uses this exact file

These workflows already load minimax_h3_ref2va_pruned_fp8_scaled.safetensors. Open one and run it as is.

- [MiniMax H3 Taomate 3-Step Acceleration | Image to Video](https://www.runcomfy.com/comfyui-workflows/minimax-h3-taomate-3-step-acceleration-image-to-video)

## Workflows: Uses another version of this file

These workflows load a different precision or format of the same model. Select minimax_h3_ref2va_pruned_fp8_scaled.safetensors in their Load Diffusion Model node to use this version.

- [MiniMax H3 Multi-Character Swap | 1-4 Characters Replacement](https://www.runcomfy.com/comfyui-workflows/minimax-h3-multi-character-image-swap-comfyui)
- [MiniMax H3 Anime Video Generator | Multi-Shot Anime + Synced Sound](https://www.runcomfy.com/comfyui-workflows/minimax-h3-in-comfyui-multi-shot-anime-synced-sound)
- [MiniMax H3 Video to Video with SolarRenderer + FastH3-Omni + Singularity V2](https://www.runcomfy.com/comfyui-workflows/minimax-h3-in-comfyui-solarrenderer-video-restyling)
- [MiniMax H3 Character Swap V1.5 with Original Audio](https://www.runcomfy.com/comfyui-workflows/minimax-h3-character-swap-v1-5-comfyui-lora-audio)
- [MiniMax H3 Character Swap with SAM3 | MDMZ](https://www.runcomfy.com/comfyui-workflows/minimax-h3-character-swap-in-comfyui-sam3-auto-prompt)
- [MiniMax H3 Motion Transfer | Side-by-Side Video](https://www.runcomfy.com/comfyui-workflows/minimax-h3-motion-transfer-in-comfyui-spatial-outpainting)
- [MiniMax H3 Performance Capture: AI Motion Transfer](https://www.runcomfy.com/comfyui-workflows/minimax-h3-performance-capture-character-swap-video)
- [VDN H3 ComfyUI Text To Video |  Faster MiniMax H3 VDN 8-Step Workflow wihout LoRA](https://www.runcomfy.com/comfyui-workflows/vdn-h3-comfyui-text-to-video-8-step-video-and-audio)
- [MiniMax H3 Latent Upscaler | BUNNYH3 Two-Pass Workflow](https://www.runcomfy.com/comfyui-workflows/minimax-h3-latent-upscaler-comfyui-two-pass-3d-video)
- [MiniMax H3 Fun Control Workflow | Controlled Image-to-Video](https://www.runcomfy.com/comfyui-workflows/minimax-h3-fun-control-in-comfyui-depth-and-pose-video)
- [MiniMax H3 8-Step Acceleration | All-in-One Video Workflow](https://www.runcomfy.com/comfyui-workflows/minimax-h3-8-step-acceleration-all-in-one-comfyui-workflow)
- [MiniMax Music Video | Full Music Video Pipeline With MiniMax H3 and Music 3](https://www.runcomfy.com/comfyui-workflows/minimax-music-video-in-comfyui-h3-music-3-workflow)
- [MiniMax H3 High Quality Dual-Sampling R2V](https://www.runcomfy.com/comfyui-workflows/minimax-h3-dual-sampling-workflow-ref2va-turbo-lora)
- [MiniMax H3 Character Replacement ComfyUI | AI Character Swap](https://www.runcomfy.com/comfyui-workflows/minimax-h3-character-replacement-comfyui-sam3-ref2va)
- [3D AI Rendering with MINIMAX H3 | 3D Rendering Video](https://www.runcomfy.com/comfyui-workflows/3d-ai-rendering-with-minimax-h3-clay-to-cinematic-video)
- [MiniMax H3 Multi References to Video 4 Steps Turbo](https://www.runcomfy.com/comfyui-workflows/minimax-h3-comfyui-4-step-reference-to-video-audio)
- [ComfyUI MiniMax H3 | Realistic Talking Avatars](https://www.runcomfy.com/comfyui-workflows/comfyui-minimax-h3-talking-avatars)
- [MiniMax H3 All-in-One | Fast Multi-Reference Video](https://www.runcomfy.com/comfyui-workflows/minimax-h3-comfyui-workflow-fast-multi-reference-video)
- [MiniMax H3 Reference to Video ComfyUI | Stereo](https://www.runcomfy.com/comfyui-workflows/minimax-h3-ref-to-video-in-comfyui-native-stereo-audio)

## Workflows: Built for reference-to-video

These workflows load another reference-to-video model in the same folder, or they are reference-to-video workflows that already have a Load Diffusion Model node. Select minimax_h3_ref2va_pruned_fp8_scaled.safetensors there.

- [Pika H3 Flex ComfyUI Workflow | 8-Step Ref2Video](https://www.runcomfy.com/comfyui-workflows/pika-h3-flex-comfyui-8-step-ref2video-stereo-audio)
- [MiniMax H3 Singularity Reference To Video ComfyUI Workflow](https://www.runcomfy.com/comfyui-workflows/minimax-h3-singularity-reference-to-video-comfyui)
- [MiniMax H3 Veda Acceleration | 3x Fast Ref-to-Video](https://www.runcomfy.com/comfyui-workflows/minimax-h3-veda-acceleration)
- [MiniMax H3 Action Scenes | Guided Action Video](https://www.runcomfy.com/comfyui-workflows/minimax-h3-action-scenes-reference-video-native-audio)
- [MiniMax H3 Long Video Generation | Motion Context V2](https://www.runcomfy.com/comfyui-workflows/minimax-h3-long-video-in-comfyui)
- [MiniMax H3 Talking & Singing Dual Character | 4-step Acceleration](https://www.runcomfy.com/comfyui-workflows/minimax-h3-comfyui-talking-singing-dual-character-4-step-acceleration)

## Same file, other names, and alternatives

- `minimax_h3_ref2va_pruned_fp8_scaled.safetensors` · This page · pruned fp8 scaled · 20.96 GB · GPUs without int8_convrot support
- `minimax_h3_ref2va_pruned_int8_convrot.safetensors` · Same model · int8 convrot · 20.97 GB · Preferred by Comfy-Org on PyTorch cu130
- `minimax_h3_ref2va_pruned_w6a8.safetensors` · Same model · w6a8 · 15.98 GB · Less VRAM
- `minimax_h3_ref2va_pruned_bf16.safetensors` · Same model · bf16 · 40.23 GB · Highest fidelity, large GPUs
- `minimax_h3_fl2va_pruned_int8_convrot.safetensors` · Sibling model · FL2VA · 20.97 GB · Text, image and first-last-frame to video

## Common errors and fixes

### Value not in list: unet_name: 'minimax_h3_ref2va_pruned_fp8_scaled.safetensors' not in [...]

Why: The workflow asks for this exact filename, but ComfyUI cannot find it in models/diffusion_models. Usually the file is missing, sits in a different folder, or was saved under another name such as minimaxH3PrunedFp8_ref2vaPrunedFp8.safetensors.

Fix: Move the file into ComfyUI/models/diffusion_models/, rename it to match the workflow (or pick it again in the dropdown), then refresh the node list or restart ComfyUI.

### torch.OutOfMemoryError: CUDA out of memory

Why: The model alone is 21 GB, before the text encoder, VAEs and activations for long clips.

Fix: Lower the resolution or clip length, enable model offloading, or switch to the w6a8 file. A 48 GB or larger cloud GPU runs it at full settings.

### SafetensorError: Error while deserializing header

Why: The download is incomplete, or the browser saved an HTML sign-in or error page under the model’s filename.

Fix: Check that the file is exactly 20,958,205,608 bytes and that its SHA256 matches the value above. Download it again from Hugging Face · Comfy-Org/MiniMax-H3 if not.

## FAQ

### Ref2VA or FL2VA, which one do I need?

Ref2VA when you want specific characters or objects from reference images to appear in the video. FL2VA for text to video, image to video, and first-last-frame to video. Many setups keep both.

### Is minimaxH3PrunedFp8_ref2vaPrunedFp8.safetensors a different model?

No. It is Civitai’s download name for this file. The SHA256 `f86f2f79ebd2d76eb8eeb46091e83982e6ff51d255747e7b16e92834b392b8e9` is the same on both sites.

### Where do I put minimax_h3_ref2va_pruned_fp8_scaled.safetensors in ComfyUI?

In `ComfyUI/models/diffusion_models/`. Then refresh the node list or restart ComfyUI so Load Diffusion Model lists the file.

### Which ComfyUI node loads minimax_h3_ref2va_pruned_fp8_scaled.safetensors?

[Load Diffusion Model](/comfyui-nodes/ComfyUI/unet-loader). It is built into ComfyUI, so no custom nodes are needed.

### Can I use minimax_h3_ref2va_pruned_fp8_scaled.safetensors commercially?

Yes, within the MiniMax H3 Community License. Commercial products must show “MiniMax H3” in their interface, and companies with more than US$20 million in yearly revenue need written permission from MiniMax first. The license also limits use to its listed territories, so read it before shipping.

## Sources

- [Comfy-Org/MiniMax-H3 on Hugging Face](https://huggingface.co/Comfy-Org/MiniMax-H3)
- [MiniMaxAI/MiniMax-H3 (original weights)](https://huggingface.co/MiniMaxAI/MiniMax-H3)
- [ComfyUI MiniMax H3 tutorial](https://docs.comfy.org/tutorials/video/minimax/minimax-h3)
- [Civitai model 2832970](https://civitai.com/models/2832970)

Canonical page: https://www.runcomfy.com/comfyui-workflows/model-files/minimax_h3_ref2va_pruned_fp8_scaled.safetensors
