realtermio/ mistral-tagger

6 Downloads Updated 2 days ago

This is a NSFW tagger

tools

ollama run realtermio/mistral-tagger

curl http://localhost:11434/api/chat \
  -d '{
    "model": "realtermio/mistral-tagger",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='realtermio/mistral-tagger',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'realtermio/mistral-tagger',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Applications

Claude Code

Claude Code ollama launch claude --model realtermio/mistral-tagger

Codex App

Codex App ollama launch codex-app --model realtermio/mistral-tagger

OpenClaw

OpenClaw ollama launch openclaw --model realtermio/mistral-tagger

Hermes Agent

Hermes Agent ollama launch hermes --model realtermio/mistral-tagger

Codex

Codex ollama launch codex --model realtermio/mistral-tagger

OpenCode

OpenCode ollama launch opencode --model realtermio/mistral-tagger

Models

Name

1 model

Size / Usage

Context

Input

mistral-tagger:latest

4.4GB · 32K context window · Text · 2 days ago

mistral-tagger:latest

4.4GB

32K

Text

Readme

mistral-tagger

A Mistral-7B-Instruct model fine-tuned via system prompt to convert natural-language scene descriptions into comma-separated Danbooru-style tags for anime/illustration image generation.

Example

Input:

A girl sitting on a bench under a cherry tree

Output:

1girl, bench, sitting, cherry tree, petals falling, soft lighting, gentle smile, school uniform, outdoors, spring, from side, wind blowing

Usage

ollama run realtermio/mistral-tagger "A boy walking in the rain"

Details

Base model: Mistral-7B-Instruct v0.3 (Q4_K_M quantization)
Context window: 32K
Output style: 8–14 tags per prompt covering subject, pose, setting, lighting, expression, clothing, atmosphere, and camera framing
Sampling: temperature 0.7, top_p 0.9 — tuned for creative variety rather than terse minimal output

Notes

This model only outputs tags — it does not generate images itself. Pair the output with your preferred image generation tool (e.g. Stable Diffusion, SDXL, etc.) that accepts Danbooru-style tag prompts.