Skip to content

JoyTag

A vision model for multi-label image tagging

Visit official site

What JoyTag is

The model is described as an AI vision model for labelling images, with a stated focus on being sex-positive and inclusive. It uses the Danbooru tag scheme for images ranging from drawings to photographs and is listed for automatic image annotation.

What you can do with it

  • Apply multi-label classification to images
  • Use the Danbooru tag scheme
  • Work with images from drawings to photographs

Application scenarios

  • Label images automatically
  • Train diffusion models that lack text pairs
  • Handle image processing needs with a sex-positive and inclusive focus

Model notes

  • The material lists classification over more than 5000 tags
  • A ViT architecture with a CNN stem and a GAP head is described
  • The material describes the model performance as strong

Pros & cons

✓ What we like

  • Covers drawings through to photographs
  • Serves automatic image annotation and diffusion model training

! What to watch out for

  • The label set follows the Danbooru tag scheme
  • Output is multi-label classification over the listed tags

FAQ

How many tags are supported?

The material states multi-label classification over more than 5000 tags.

What architecture is used?

The material describes a ViT architecture with a CNN stem and a GAP head.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review