Three Avatar Styles, One Person (Chibi, Flat Illustration, Semi-Realistic)

【用途】把同一個人物，一次畫成三種不同風格的圓形頭像（同一張圖、左中右並排），方便挑哪種風格當社群頭貼／團隊大頭照，尺寸 [尺寸比例]。

【人物設定｜三格是同一個人，特徵要對得上】
人物：[人物描述]
要保留的識別特徵（三格都要看得出來是同一人）：[識別特徵]
表情與視角：[表情視角]

【三格風格｜由左到右】
- 左：Q 版可愛風……

[ Log in to see the full prompt ]

This is a scouting prompt for when you're not sure which avatar style suits you — seeing all three styles side by side in one image is far less work than three separate back-and-forth generations. The hard part is keeping it 'the same person,' so the identifying features are pulled out into their own variable, and the hard rules require the facial features to match across all three panels — this is the practical technique for getting GPT Image 2 to hold character consistency. Two constraints are deliberate: 'illustration, not a photo' prevents the semi-realistic panel from sliding into an uncanny photorealistic look, and 'do not depict a real celebrity' is a hard line for likeness rights and content safety — describe the character with generic features like 'a short-haired woman wearing round glasses' rather than naming a real celebrity. The circular crop and solid-color background are there so the result drops straight into a social profile picture frame.