# MiniMax H3 INT8 Reference-to-Video

## Summary

`MiniMax H3` reference-to-video workflow from the same-day `javawock7618` pack. This is the more direct character-consistency path.

## 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/resolve/main/MiniMax_int8_R2V-javano2608.7.json>
- Original publisher: `javawock7618`
- Date imported: `2026-08-09`

## Developer notes

- The source README describes this as the `Reference-to-Video` workflow.
- It keeps the same INT8, Sol-Attn, Spectrum, Lightx2v, Turbo LoRA, and latent-upscaler stack as the FLF workflow.
- Motion Context is not part of this one, which may make it the cleaner baseline.

## Our notes

- This is the MiniMax import most worth comparing against LTX reference-control workflows.
- It is still heavy for an `RTX 3070`, but it is the most practical way in this pack to test whether MiniMax H3 is becoming more than a curiosity.
- Good choice when you care more about reference retention than keyframe choreography.

## Required custom nodes

- `ComfyUI-SolAttn_triton`
- `ComfyUI-Spectrum-MiniMax-H3`
- `ComfyUI-H3-Motion-Context` is not used here
- `ComfyUI-MiniMaxH3_LatentUpscaler`

## Required models

- `Comfy-Org/MiniMax-H3`
- `Kijai/MiniMax-H3_comfy`
- `drbaph/MiniMax-H3-Turbo-Lora-ComfyUI` if the turbo path is enabled

## Notes

- Best same-day MiniMax workflow for reference retention.