ComfyUI Node
MiniMax H3 Prompt Writer ⚡
Write a MiniMax H3 prompt with a local text encoder. The model supplies the three core fields; the node assembles the exact MiniMax envelope around them.
MiniMax H3 Prompt Writer ⚡
- clip
- prompt
- raw_output
◄text_encoder▾►
◄ideaA lighthouse keeper climbs the stairs at dawn and looks out over a calm sea.►
◄modet2va►
◄length243 (~10.13s)►
◄seed0►
◄max_length768►
◄temperature0.70►
◄include_musictrue►
Category🤖 CCTech/MiniMax H3
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| text_encoder | COMBO | Local LLM that writes the prompt. Needs a generation-capable encoder such as Gemma-3 12B or a Qwen3-VL. The architecture is detected from the file; ignored when a clip is connected. | |
| idea | STRING | A lighthouse keeper climbs the stairs at dawn and looks out over a calm sea. | Your raw idea. Everything you state here is kept; anything vague is filled in with concrete detail. |
| mode | COMBO | t2va | t2va: text only. i2va: from a first frame. fl2va: between a first and last frame. l2va: landing on a last frame. |
| length | COMBO | 243 (~10.13s) | Clip length. Sets the reference-alignment time and the latest valid shot cut. |
| seed | INT | 00–18446744073709550000 | — |
| max_length | INT | 76864–8192 | — |
| temperature | FLOAT | 0.700–2 | 0 generates deterministically. |
| include_music | BOOLEAN | true | Off writes non_diegetic_music: N/A, for clips that should carry no score. |
| clipopt | CLIP | Optional. Overrides the dropdown and avoids reloading, e.g. the CLIP already loaded for the video model. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| prompt | STRING | — |
| raw_output | STRING | — |