Downloads · 30 days
2.4K
17% of all-time downloads
minimax_h3_ref2va_fp8_e4m3fn
rzgar/minimax_h3_ref2va_fp8_e4m3fn
minimax_h3_ref2va_fp8_e4m3fn is a image-to-video model from rzgar. Use it for the image-to-video task on the model card, and read the license before you ship it in a product. It is set up for diffusers. The card lists the license as apache-2.0.
Quantized versions of the MiniMax H3 Ref2VA (reference-to-video-audio) diffusion model. Both keep the text-conditioning path (tokenrefiner, conditioning projections, audio heads) in full precision.
All-time downloads
13.5K
Public
Repo size
76.3 GB
Likes
5
Trending 1
Repo makeup
Click a slice to open those files.
.safetensors76.3 GB · 100%
At a glance
- Task
- Image-to-Video
- Library
- diffusers
- License
- apache-2.0
- Access
- Public
- Created
- Aug 7, 2026
- Updated
- Aug 8, 2026
- SHA
- f6888797
Base models
Model card
From the Hugging Face model README
MiniMax H3 Ref2VA (mxfp8 & fp16attn_fp8)
Quantized versions of the MiniMax H3 Ref2VA (reference-to-video-audio)
diffusion model. Both keep the text-conditioning path (token_refiner,
conditioning projections, audio heads) in full precision.
Latest comfy version improves audio quality at low steps.
minimal 5 step (talking heads, less motions) | balanced 8-steps (shift 8) | Action-rich scenes above 12 (shift max 12)
<video controls width="1280" height="720"> <source src="https://huggingface.co/rzgar/minimax_h3_ref2va_fp8_e4m3fn/resolve/main/demo/demo_comfyV031x.mp4" type="video/mp4"> Your browser does not support the video tag. </video> <img src="https://huggingface.co/rzgar/minimax_h3_ref2va_fp8_e4m3fn/resolve/main/demo/ksampler.png" width="1280" height="720" />| File | Download | Note |
|---|---|---|
minimax_h3_ref2va_fp16attn_fp8.safetensors | Download 28.7GB | FP16 attention + FP8-E4M3 MLP |
minimax_h3_ref2va_mxfp8.safetensors | Download 47.6GB | E4M3 weights, E8M0 block scales |
Comparison
| fp16attn_fp8 | mxfp8 | |
|---|---|---|
| Size | 28.7 GB | 47.6 GB |
| Base checkpoint | pruned fp8_scaled (curve-form, adaln_t_table) | full bf16 (time-embedder) |
| Attention (qkv/out) | FP16 (exact) | MXFP8 (block-quantized) |
| MLP (fc1/fc2) | FP8-E4M3 per-tensor absmax | MXFP8 block-quantized |
| Quant scale granularity | per-tensor | per-32 block (better outlier handling) |
input_scale | calibrated (fc1), fpmm on fc2 | 1.0 identity (150 layers) |
| token_refiner / conditioning | full precision | full precision |
| Relative precision | higher than pruned fp8, attn lossless | highest of the two (finer scales) |
| Memory footprint | lighter (faster load, less streaming) | heavier |
| Best for | everyday use / speed-first setups | maximum quality per weight bit |
- Task
- Image-to-Video
- Library
- diffusers
- License
- apache-2.0
- Created
- Aug 7, 2026
- Updated
- Aug 8, 2026