danbot-comfy-node
A ComfyUI extension with 13 custom nodes.
Nodes (13)
The typo in the name hides a genuinely useful second pass
The sampler settings for a text model, not an image model
Where a sentence becomes a tag list
A pre-built 'do not say these tags' list for the tag model
Load a real tag-translation model, not an API call
Split the model's tag output into copyright, character, and everything else
Join up to six strings with a separator you control
The debugging node that shows you what the tag model actually said
A plain multiline text box that outputs a string — nothing more
Turn any image size into the aspect-ratio tag the model expects
The node that builds the model's actual prompt string
The one node that does the whole translate-and-extend dance
The rating, length, and aspect dials for the tag model
danbot-comfy-node
Danbooru Tags Translator for ComfyUI. This custom node allows you to convert natural language prompts in Japanese and English into Danbooru tags.
Installation
Manual Installation
# activate your python env
cd custom_nodes
git clone https://github.com/p1atdev/danbot-comfy-node
cd danbot-comfy-node
pip install -r requirements.txt
If you are using portable ComfyUI, use your/ComfyUI_windows_portable/python_embeded/python.exe -s -m pip intead of pip. For example:
cd custom_nodes
git clone https://github.com/p1atdev/danbot-comfy-node
cd danbot-comfy-node
../../../python_embeded/python.exe -s -m pip install -r requirements.txt
ComfyUI Registry
comfy node registry-install danbot-comfy-node
Example Workflows
See workflows directory.
| Filename | Description | | - | - | | Translation+Extension.json | Example workflow of translating and extending Danbooru tags from a Japanese prompt. | | Translation+Extension+Image_Generation.json | Example workflow of translating and extending Danbooru tags from a Japanese prompt, and then generating an image using AnimagineXL 4.0. |
Example Outputs
| Input prompt | Translated tags | Extended tags | Generated image |
| - | - | - | - |
| 猫耳で黒髪ロング、制服を着ており、目は黄色の少女。背景はハーフトーンのついた青で、白枠が付いている。ソファーに座って足を組みながらこっちを見ている。 | 1girl, solo, blue background, halftone background, looking at viewer, animal ears, school uniform, yellow eyes, black hair, long hair, sitting, crossed legs, cat ears, border, halftone, white border, couch | shirt, skirt, closed mouth, very long hair, short sleeves, white shirt, full body, black skirt, pleated skirt, black footwear, collared shirt, socks, black socks, outside border | <img src="./assets/outputs/sample_01.jpg" width="600px" /> |
| ピクセルアート。猫耳の女の子がダンボール箱に入っている。chibi。青と白の二色の髪色。パーカーを着ている。 | 1girl, multicolored hair, blue hair, white hair, two-tone hair, chibi, cat girl, hoodie, hood, box, pixel art, in container, in box | solo, long sleeves, animal ears, very long hair, sleeves past wrists, sleeves past fingers, blue eyes, long hair, full body, cat ears, chibi only, tail, cat tail, cardboard box | <img src="./assets/outputs/sample_02.jpg" width="600px" /> |
| VOICEVOXのずんだもんと東北きりたんが東京の街中で立っている。写真の背景。全身が写ってる。笑顔。 | voiceroid, voicevox, touhoku kiritan, zundamon, 2girls, multiple girls, photo background, smile, standing, full body, road, street | photo background, photo-referenced, open mouth, shirt, long sleeves, closed mouth, orange eyes, brown hair, green hair, ahoge, short sleeves, medium hair, white shirt, hands on own hips, looking at another, black footwear, half-closed eyes, closed eyes, t-shirt, layered sleeves, short over long sleeves, japanese clothes, kimono, sash, socks, white socks, green shorts, shorts, white kimono, headgear, obi, lamppost, green overalls | <img src="./assets/outputs/sample_03.jpg" width="600px" /> |
- Prompt generation
- Translation
- rating:
general - length:
very_short - template_name:
translation
- rating:
- Extension
- rating:
general - length:
long - template_name:
extension
- rating:
- Generation config
- max_new_tokens:
256 - do_sample:
true - temperature:
1.00 - top_p:
1.0 - top_k:
50 - min_p:
0.05 - num_beams:
1
- max_new_tokens:
- Seed: 347414205
- Translation
- Image generation
- Image model: AnimagineXL 4.0 opt
- Prompt suffix (quality tags):
masterpiece, best quality, high score, great score, latest - Negative prompt:
lowres, bad anatomy, bad hands, text, error, missing finger, extra digits, fewer digits, cropped, worst quality, low quality, low score, bad score, average score, signature, watermark, username, blurry, - Image size: 1024x1024
- Seed:
944162813372176 - Steps:
25 - CFG:
5.0 - Sampler name:
euler_ancestral - Scheduler:
normal - Denoise:
1.00
Available models
| Model name | Knowledge cutoff | Param size | | - | - | - | | 🤗 DanbotNL 2408 260M| 2024/8/31 | 262M |