# MiniMax H3 Face Refine Javawock Pack

## Summary

Same-day `MiniMax H3` face refinement workflow from the `javawock7618` pack. This is the clearest new MiniMax H3 path for fixing weak or distant faces.

## Model Family

- `minimax`

## Status

- `imported-only`

## Source

- Workflow pack page: <https://huggingface.co/javawock7618/comfy-MiniMax-H3-workflows>
- Workflow file: <https://huggingface.co/javawock7618/comfy-MiniMax-H3-workflows/raw/main/MiniMax_int8_FR-javano2608.3.json>
- Original publisher: `javawock7618`
- Date imported: `2026-08-22`

## Developer notes

- The source README describes this as the `Face Refine` workflow.
- It uses the same MiniMax H3 INT8 base stack as the other workflows in the pack.
- The graph includes a dedicated face refinement stage and notes that small faces may need a larger crop size.

## Our notes

- This looks like a real fix candidate for MiniMax H3 face problems, not just a cosmetic polish pass.
- Best new same-day item in the pack if you care about faces at distance or identity cleanup after generation.
- Still not cheap on a `3070`, but the focus is narrow enough that it is worth keeping around.

## Required custom nodes

- `ComfyUI-H3-FaceRefine`
- `ComfyUI-KJNodes`
- `ComfyUI-VideoHelperSuite`
- `ComfyUI-SolAttn_triton`
- `ComfyUI-Spectrum-MiniMax-H3`

## Required models

- `minimax_h3_fl2va_pruned_int8_convrot.safetensors`
- `minimax_h3_video_vae_int8_convrot.safetensors`
- `minimax_h3_audio_vae_fp32.safetensors`
- `qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors`
- `H3/minimax_h3_ref_lora_rank_256_bf16.safetensors`
- `H3/minimax_h3_ref2v_lightx2v_turbo_4step_v0.1_resized_avg_rank_20_bf16.safetensors`

## Notes

- Best same-day MiniMax workflow for face cleanup.