ComfyUI Node
⚗ Alchemica Prompt Enhancer (Krea 2 text encoder)
Prompt enhancer that uses Krea 2's own text encoder: no Ollama, it can also read images.
⚗ Alchemica Prompt Enhancer (Krea 2 text encoder)
- clip
- image
- scene
- detail
- negative
- report
◄idea►
◄stylerealistic photo►
◄words80►
◄creativity0.00►
◄seed0►
◄skinsmooth (no body hair)►
◄instructions►
Category⚗ AlchemicaMente/conditioning
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| clip | CLIP | Krea 2's text encoder: it writes the prompt, then encodes it | |
| idea | STRING | Any language, a few words are enough. Empty + image = describes the image | |
| style | COMBO | realistic photo | 8 options: realistic photo, portrait, still life / lab, landscape, architecture, cinematic, +2 |
| words | INT | 8030–200 | Length of the SCENE (the detail is about 30-50% of it) |
| creativity | FLOAT | 0.000–1.5 | 0 = deterministic (same text every run). 0.6-0.9 = variations; change the seed |
| seed | INT | 00–18446744073709550000 | Only matters when creativita (creativity) > 0 |
| imageopt | IMAGE | Visual reference: the model looks at it (Qwen3-VL is a vision-language model) | |
| skinopt | COMBO | smooth (no body hair) | Skin rendering in the DETAIL. smooth = no body/vellus hair (also added to the negative); natural = fine vellus hair (the 1.0.0 behaviour); model decides = hair not mentioned |
| instructionsopt | STRING | Your own extra instructions for the enhancer, any language. They are added to its rules and win when they conflict. E.g. 'Skin: smooth, tanned. Always golden hour light.' |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| scene | STRING | — |
| detail | STRING | — |
| negative | STRING | — |
| report | STRING | — |