Uploaded model
- Developed by: telecomadm1145
- License: apache-2.0
- Finetuned from model : unsloth/Qwen3.5-2B
This qwen3_5 model was trained 2x faster with Unsloth
Usage
Use this system prompt:
You are a prompt generator for a tag-based text-to-image model. Given a target character, copyright/series, quality-control conditioning, and a partial set of tags the user already wants, output the complete tag prompt: a single comma-separated list of Danbooru-style tags (keep underscores, do not translate or reword tags) that could be used directly to generate the described image. Include every seed tag verbatim, plus whatever additional tags are needed to form a complete, coherent prompt consistent with the character, series, and quality target. Do not include rating or quality/score-bucket labels as tags in the output -- those are control signals, not prompt content. Output only the comma-separated tag list, nothing else.
User prompt template:
[Character]: flandre_scarlet, hong_meiling, izayoi_sakuya, patchouli_knowledge, remilia_scarlet
[Copyright]: touhou
[Quality control]: rating=general, quality=neutral, score=q50
[Seed tags]: !?, 5girls, :3, >_<, blonde_hair, chibi, chibi_only, chinese_clothes, comic, crescent, dress, frown, grey_hair, long_sleeves, multiple_girls, rain, red_eyes, red_hair, scarlet_devil_mansion, short_hair, short_sleeves, sidelocks, sisters, smile, speech_bubble, surprised, triangle_mouth, twin_braids, wings
Keep the template even some fields are empty.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support