Can the output be used as a Live2D base?

Yes for the visual reference. Live2D itself requires manual layer separation by a designer — the AI image is the starting point, not the final rigged file.

How do I make a matching expression sheet?

Lock the seed and re-generate with different expression keywords ('smiling', 'surprised', 'sad', 'laughing'). You'll get 4 consistent variants.

Prompt Recipe

VTuber Avatar AI Prompt — Full Body Reference

Frontal standing pose + neutral background + costume detail — the Live2D-ready base

このプロンプトで狙う作風の参考画像

プロンプト

コピーして他の AI ツールで使うことも、下の CTA で Studio に自動入力することもできます。

vtuber standing portrait, anime style, simple clean background, full body visible, clear line art, VTuber character design, full body, standing frontal pose, looking at viewer, neutral expression, detailed asymmetric costume, vibrant hair color with highlights, neutral light background, character design sheet style, soft cinematic lighting, masterpiece, highly detailed

ネガティブプロンプト

lowres, blurry, deformed, extra limbs, cropped, dynamic pose, watermark

このイラストを生成

登録だけで 100 クレジット（約 25 枚分）無料付与

推奨設定

モデル: FLUX.2
解像度: 768×1024 (3:4 縦・標準)
1 生成: ≈ 4 cr
Steps: 20
CFG: 3.5

使用スタイル: FLUX.2 — 汎用の高品質画像モデル。スタイルはプロンプトで指定（アニメ／実写／水彩など自在）

コツ

Frontal symmetric pose is critical for Live2D rigging — avoid dynamic poses for the base reference
Distinctive features (animal ears, horns, unique accessories) help your VTuber stand out
Use ComfyStudio's background-removal workflow afterward to get a transparent PNG ready for streaming overlay

よくある質問

Q. Can the output be used as a Live2D base?: A. Yes for the visual reference. Live2D itself requires manual layer separation by a designer — the AI image is the starting point, not the final rigged file.
Q. How do I make a matching expression sheet?: A. Lock the seed and re-generate with different expression keywords ('smiling', 'surprised', 'sad', 'laughing'). You'll get 4 consistent variants.