ComfyUI Node
LC Vision Danbooru Caption 🏷️
Prompts for SDXL, Pony and Illustrious: a short caption, a detailed caption, Danbooru tags, or tags plus a sentence, from one image or from a rough prompt. Short enough for SDXL, with optional quality tags and a matching negative.
LC Vision Danbooru Caption 🏷️
- vision_model
- image
- positive
- negative
◄modeImage analysis►
◄outputDanbooru tags►
◄prompt►
◄negative►
◄style_tagNone►
◄quality_tagsnone►
◄max_tokens150►
◄temperature0.30►
◄seed0►
CategoryLC Vision
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| vision_model | LC_VISION_MODEL | Handle from LC Vision Loader. | |
| mode | COMBO | Image analysis | Image analysis: writes the prompt from the image. Prompt enhance: rewrites the idea written in prompt. |
| output | COMBO | Danbooru tags | short caption (80 tokens max), detailed caption (150), Danbooru tags (150), prompt gen mixed: tags plus a sentence (200). Each finetune likes a different one: Pony and Illustrious lean on tags, realism finetunes on captions. |
| prompt | STRING | Prompt enhance: the rough idea to rewrite. Not used in Image analysis. | |
| negative | STRING | Your own negative prompt, used in both modes. It goes first in the negative output, followed by the automatic anatomy / artifact / censorship tags (duplicates removed). | |
| style_tag | COMBO | None | Commits the prompt to a visual style. 'None' leaves it as the image or idea is. |
| quality_tags | COMBO | none | Adds the quality / score tags up front and their opposites to the negative. They sometimes do more harm than good: off by default. |
| max_tokens | INT | 15016–1024 | How long the prompt can be, for every output. The instructions follow it (about max_tokens / 4 tags, captions about 0.7 words per token). Rough guide: short caption 80, detailed caption 150, tags 150, mixed 200. |
| temperature | FLOAT | 0.300–2 | Low = literal and repeatable. Tags want it low. |
| seed | INT | 00–4294967295 | — |
| imageopt | IMAGE | The picture to describe (Image analysis). The first frame of a batch is used. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| positive | STRING | The prompt for your SDXL / Pony / Illustrious text encoder. |
| negative | STRING | A matching negative prompt (with the score / quality tags when quality_tags is on). |