A custom node for ComfyUI that makes writing MiniMax H3 reference prompts much faster and more readable.
Instead of typing long tags like <Picture 1>, <Subject 2>, <Video 3>, <Audio 4> by hand, you type short shorthands like @Pic 1, @Sub 2, @Vid 3, @Aud 4. The node converts them to the real reference tags that the MiniMax H3 tokenizer understands, at execution time — so the model still receives the exact same conditioning.
@autocomplete menu — while typing in the text area, press@to open a menu of the available tag kinds, then pick a number.@+ letter hotkeys — type@thenV/P/S/Ato immediately insertVid/Pic/Sub/Audand open the number picker.- Number menu — after choosing a tag kind, a second menu lets you pick the index (Vid 1–3, Pic 1–9, Sub 1–12, Aud 1–3).
- Yellow + underline highlighting — shorthand tags are highlighted in bright yellow with an underline so they stand out at a glance.
- Paste auto-conversion — paste prompt text containing full tags (
<Picture 1>,<Video 2>,<Subject 3>,<Audio 4>) and they are automatically rewritten to the short@-form. - No space required after the number — tags are recognized even when directly followed by Chinese or English text (e.g.
@Pic 1中文提示词→<Picture 1>中文提示词). - Smart digit handling — tag indices are capped to each tag type's valid range, so text like
@Pic 912is parsed as<Picture 9>12(the12stays as normal prompt text). - Output stays correct — the node always emits the real
<Picture 1>/<Video 2>/<Subject 3>/<Audio 4>tags downstream.
| Shorthand | Output tag | Valid indices |
|---|---|---|
@Pic N |
<Picture N> |
1 – 9 |
@Vid N |
<Video N> |
1 – 3 |
@Sub N |
<Subject N> |
1 – 12 |
@Aud N |
<Audio N> |
1 – 3 |
The @ is optional — Pic 1 works the same as @Pic 1.
These tag limits match the MiniMax H3 model: up to 12 total reference assets, with ≤9 images, ≤3 videos, and ≤3 audio inputs.
-
Copy this folder into your ComfyUI
custom_nodesdirectory:cp -r ComfyUI-H3PromptBooster /path/to/ComfyUI/custom_nodes/
The folder must be placed so that the plugin's
__init__.pyis directly insidecustom_nodes/:custom_nodes/ └── ComfyUI-H3PromptBooster/ ├── __init__.py ├── nodes.py └── web/ └── js/ └── h3PromptTags.js -
Restart ComfyUI. The plugin auto-loads on startup — there is no manual step.
-
In the node search bar, look for "H3 Prompt Booster" (category
prompt/minimax).
- Add the H3 Prompt Booster node to your workflow.
- Connect its STRING output to the
promptinput of a MiniMax H3 node (e.g.MiniMaxH3ReferenceToVideo). - Type your prompt, using
@Pic 1/@Vid 2/@Sub 3/@Aud 4(or use the@menu). - The node outputs the real
<Picture 1>/<Video 2>/<Subject 3>/<Audio 4>tags.
- ComfyUI (recent version, tested on 0.30.x)
- Python 3
- No additional Python packages required.