Nodes/MOSS漫画コマ割りデータセット/Character appearance (image / text)
ComfyUI Node

Character appearance (image / text)

Pass a character's appearance as text, or extract it from a reference image with the MOSS Qwen.

By sepiablue-ai·Created 21 days ago·Updated 18 days ago· 43
Character appearance (image / text)
  • Qwen
  • 参考画像
  • Appearance for shared characters
◄指定方法文章のみ►
◄外見プロンプト►
◄最大トークン数256►
CategoryManga panel dataset

Inputs (5)

NameTypeDefaultDescription
指定方法COMBO文章のみ"Text only" passes the appearance prompt as is. The other methods extract the appearance from the reference image.
外見プロンプトSTRINGFor "Text only", describe the appearance in English. With an image, you can add details such as hair color or clothing. Put pose and background in the panel's scene.
最大トークン数INT25664–1024Maximum length of the extracted appearance text.
QwenoptCLIPYou can share the CLIPLoader (type=moss_image) output used for MOSS generation. Not evaluated for "Text only".
参考画像optIMAGEConnect one image that shows the character from a Load Image node. If needed, it is downscaled to 768 px on the long side for analysis only.

Outputs (1)

NameTypeDescription
Appearance for shared charactersSTRING—