42 lines
1.4 KiB
Markdown
42 lines
1.4 KiB
Markdown
# MiniMax H3 Prompt Rewriter 8B ComfyUI
|
|
|
|
## Summary
|
|
|
|
ComfyUI workflow bundle for turning raw MiniMax H3 ideas into the structured prompt format used by the prompt rewriter model.
|
|
|
|
## Model Family
|
|
|
|
- `minimax`
|
|
|
|
## Status
|
|
|
|
- `imported-only`
|
|
|
|
## Source
|
|
|
|
- Model card: <https://huggingface.co/Codes4Fun/lightx2v_MiniMax-H3-Prompt-Rewriter-8B_ComfyUI>
|
|
- Workflow file: <https://huggingface.co/Codes4Fun/lightx2v_MiniMax-H3-Prompt-Rewriter-8B_ComfyUI/resolve/main/comfyui-minimax-h3-rewriter-8b.json>
|
|
- Source README notes that this is a merge of `lightx2v/MiniMax-H3-Prompt-Rewriter-LoRA-8B` with `Qwen/Qwen3-VL-8B-Instruct`
|
|
- Date imported: `2026-08-21`
|
|
|
|
## Developer notes
|
|
|
|
- The workflow is a helper, not a generator. It rewrites prompts for MiniMax H3-style video and audio/video tasks.
|
|
- The source graph includes a `TextGenerate` node plus a `CLIPLoader` wired to `qwen3vl_8b_rewriter_bf16_comfyui.safetensors`.
|
|
- Most of the rest of the graph looks like stock ComfyUI plumbing.
|
|
|
|
## Our notes
|
|
|
|
- Useful if you want cleaner MiniMax prompts without hand-tuning the long format every time.
|
|
- Less interesting than a direct generation workflow, but still practical and easy to slot into a MiniMax H3 prep chain.
|
|
|
|
## Required custom nodes
|
|
|
|
- MiniMax-compatible `TextGenerate` node from the source workflow
|
|
- standard ComfyUI core nodes used by the graph
|
|
|
|
## Notes
|
|
|
|
- Good helper workflow for prompt preparation.
|
|
- Not a standalone video model.
|