Home Creators Posts Import Register Favorites Logout

Downloads

  • hs-anima-3.0.zip

Some files exceed the archive size limit and weren't saved: 1 file. Please note these limits are in place to keep this site running long term, without costing a fortune. You can favorite the creator though, certain milestones increase the limit.

Content

如果您是新手,非常推薦直接將我的作品圖片拖入 ComfyUI 中「抄作業」學習!

If you're a beginner, it's highly recommended to just drag and drop my sample images directly into ComfyUI to copy the workflow!

hs-anima-3.0 模型說明

hs-anima-3.0 是基於 Anima 深度微調的衍生模型。模型全面支援 2D、3D 與寫實等多元風格,相容各類型 LoRA,並對 hshiKoi 系列 LoRA 進行了專屬適配優化。

繪師標籤與提示詞規範

模型繼承並對齊了底層 Anima 的標籤體系,調用繪師與風格時請遵循以下規則:

繪師標籤格式:繪師名稱前必須加上 @ 前綴(例如 @artist_name),否則風格效果會非常微弱。

標籤書寫格式:使用全英文小寫,單字間以空格取代底線(僅品質評分標籤保留底線,如 score_7)。

標籤順序結構:

[品質/年份/安全標籤] [主體數量] [角色名] [作品名] [@繪師名] [一般特徵與風格標籤]

自然語言混用:支援標籤與英文自然語言混合輸入;若需要精準呈現角色特徵與畫面細節,建議使用至少 2 句以上的具體描述。

LoRA 相容性

全類型相容:全面相容各類型角色、服裝、構圖與風格 LoRA。

hshiKoi 專屬適配:對 hshiKoi 系列模型與 LoRA 具備最高相容性與融合表現。

hs-anima-3.0 Model Card

hs-anima-3.0 is a fine-tuned model based on Anima. It supports a comprehensive range of visual styles—including 2D, 3D, and realism—features universal LoRA compatibility, and is specially optimized for hshiKoi LoRAs.

Artist Tags & Prompting Guidelines

hs-anima-3.0 follows Anima's core tagging conventions:

Artist Tag Syntax: You must prefix artist names with @ (e.g., @artist_name). Omitting @ will result in very weak style adherence.

Tag Formatting: Use lowercase letters and spaces instead of underscores (underscores are only used for score tags like score_7).

Tag Order:

[quality/year/safety tags] [subject count] [character] [series] [@artist] [general features & style tags]

Natural Language: Fully supports mixing Danbooru-style tags with natural language captions. Detailed descriptions (at least 2 sentences) are recommended for precise character and scene rendering.

LoRA Compatibility

Universal Compatibility: Compatible with all types of character, outfit, pose, and style LoRAs.

Optimized for hshiKoi: Features the highest compatibility and optimal blend performance with the hshiKoi model and LoRA ecosystem.

Files

Comments

No comments found for this post.