动漫寿司吧场景转换
GPT Image 2 提示词GPT Image2提示词GPT Image 2 提示词案例AI提示词案例AI图片生成提示词案例AI绘画提示词案例OpenAI图像生成提示词ChatGPT画图提示词
Full GPT Image 2 Prompt
Using REFERENCE_0 as the artistic base, transform the scene from a colorful fantasy market into a warm, detailed Japanese sushi restaurant interior while keeping the same ornate storybook-anime watercolor feeling, dense linework, glowing lights, and cozy golden atmosphere.
Goal: Recompose the two-girl scene into an intimate sushi-counter moment.
Main changes:
- Replace the market stalls and floating colored bubbles with a traditional wooden sushi bar filled with hanging lanterns, shelves of dishes, wall scrolls, patterned tiles, noren curtains, and chefs working behind the counter.
- Keep exactly 2 anime girls, but change their roles and styling:
1. Left girl: a {argument name="left character role" default="green-haired sushi waitress"} leaning forward from behind the counter and presenting food.
2. Right girl: a {argument name="right character role" default="long pale-blue-haired customer"} seated at the counter in an elegant icy-blue kimono-style dress, watching the food.
- Change the cuisine to {argument name="cuisine type" default="sushi"}. Add a small wooden serving board with exactly 5 visible sushi pieces: red tuna nigiri, pale white fish nigiri, yellow egg nigiri, salmon nigiri, and one seaweed-wrapped gunkan piece with orange topping.
- Add tabletop details: one blue floral teacup, one small soy sauce dish, one black rectangular bento tray with exactly 3 compartments, and small illustrated menu cards.
- Add background staff and diners as soft supporting figures, not the focus.
- Apply {argument name="face treatment" default="flat opaque beige rectangular blocks covering the visible faces"} to the main girls and background people, as if the face areas are intentionally masked.
Style constraints: Preserve the highly detailed hand-drawn watercolor/colored-pencil texture, luminous warm lighting, decorative pattern density, soft fantasy-anime atmosphere, and slightly nostalgic DALL-E-3-like illustration quality. Avoid photorealism, avoid modern minimalism, and do not add extra main characters.