PROMPT RECORD图像记录

Seedance 中真正能用的角色表提示词。🎬 Higgsfield 刚刚把他们那部 200 万美元 AI 电影的幕后全部开源，而最大的心得就是关于角色表：3 个面板（正面、背面、面部特写）—...

Nano Banana meAsifAi Fri Sep 04 15:43:00 +0000 2026

The character sheet prompt that actually holds up in Seedance. 🎬

Higgsfield just open-sourced everything from their $2M AI feature film, and the biggest lesson was about character sheets: 3 panels (front, back, close portrait) — and TWO close-ups, one neutral and one smiling, or the model invents someone else's teeth.

I turned that into a master prompt for Nano Banana Pro on Higgsfield. Upload one photo of any person, paste the prompt, and you get a 4-panel sheet:

🧍 Front full body — head to shoes, hands visible
🔄 Back full body — same pose, same scale
😐 Close portrait, 3/4 view, neutral
😄 Close portrait, 3/4 view, genuine smile with real teeth

Same face, same outfit, same hair, same proportions in every panel. Neutral grey backdrop, even 5600K studio light, print-ready.

Works for male or female — the prompt adapts grooming and proportions automatically.

Tips:
* Use a clear front-facing photo with even lighting, no filters
* Set aspect ratio to 16:9 so four panels fit
* Pro trick from Higgsfield's own film: crop the head out of the two full-body panels so the model only copies the face from the close-ups
* Feed the finished sheet into Soul ID and your character stays locked across every scene

Prompt is below 👇

Create a professional character sheet based strictly on the attached reference photo. The sheet is one single image with four panels side by side on a plain neutral grey studio backdrop, separated by thin dark dividers, presented as a technical model turnaround with the exact realistic photographic style of the reference.

The subject is the same real person as in the attached photo, with the identical face, facial structure, skin tone, eye color, eyebrows, nose, lips, hairstyle, hair color, facial hair, glasses if present, age, height and body build. If the person is male, keep his male grooming and proportions exactly; if female, keep her female features, hairstyle and proportions exactly. Outfit: keep the same clothing as in the reference photo, with real fabric texture and natural creases; all clothing is opaque, closed and fully covering. If the reference only shows the upper body, complete the outfit with plain dark trousers and simple shoes that match the top.

Panel 1 — FRONT, full body: full-length front view, head to shoes, standing straight and neutral facing the camera, arms relaxed at the sides with both hands clearly visible and intact, feet visible, even framing with headroom and foot room, the whole outfit visible.

Panel 2 — BACK, full body: full-length back view, same standing pose, facing directly away from the camera, showing the back of the hair, the back of the outfit and the shoes from behind, same scale and alignment as panel 1.

Panel 3 — CLOSE PORTRAIT, neutral: large close-up portrait from the chest up in three-quarter view, face turned about 30 degrees, eyes looking toward the lens, calm neutral expression with lips closed, sharp detail on every facial feature, natural skin texture with pores, readable eye catchlights.

Panel 4 — CLOSE PORTRAIT, smiling: the same three-quarter close-up framing as panel 3, now with a genuine warm smile showing the person's real teeth, eyes slightly crinkled, same head angle and lighting.

Lighting: neutral soft even wrap-around studio light, near-shadowless, daylight-balanced 5600K, no color cast, identical across all four panels. Sharp realistic photograph, crisp and print-ready.

Locks: one identical person in all four panels — same identity, same outfit, same hair, same proportions; feet visible in both full-body panels; identical backdrop, scale and light in every panel; photoreal, no illustration or 3D look.

Seedance 中真正能用的角色表提示词。🎬 Higgsfield 刚刚把他们那部 200 万美元 AI 电影的幕后全部开源，而最大的心得就是关于角色表：3 个面板（正面、背面、面部特写）——再加上两个特写，一个中性表情，一个微笑，否则模型就会凭空编造出别人的牙齿。 我把它改造成了 Higgsfield 上 Nano Banana Pro 的万能提示词。上传任意一个人的照片，粘贴提示词，就能得到一张 4 面板的角色表： 🧍 正面全身 —— 从头到鞋，双手可见 🔄 背面全身 —— 同样姿势，同样比例 😐 面部特写，四分之三侧脸，中性表情 😄 面部特写，四分之三侧脸，露出真实牙齿的真诚微笑 四张图中同一张脸、同一套衣服、同样的发型、相同的身体比例。中性灰色背景，均匀的 5600K 影棚光，可直接印刷。 无论男女都...

一键生图查看来源

中文说明

Seedance 中真正能用的角色表提示词。🎬 Higgsfield 刚刚把他们那部 200 万美元 AI 电影的幕后全部开源，而最大的心得就是关于角色表：3 个面板（正面、背面、面部特写）——再加上两个特写，一个中性表情，一个微笑，否则模型就会凭空编造出别人的牙齿。我把它改造成了 Higgsfield 上 Nano Banana Pro 的万能提示词。上传任意一个人的照片，粘贴提示词，就能得到一张 4 面板的角色表： 🧍 正面全身 —— 从头到鞋，双手可见 🔄 背面全身 —— 同样姿势，同样比例 😐 面部特写，四分之三侧脸，中性表情 😄 面部特写，四分之三侧脸，露出真实牙齿的真诚微笑四张图中同一张脸、同一套衣服、同样的发型、相同的身体比例。中性灰色背景，均匀的 5600K 影棚光，可直接印刷。无论男女都...

原始 Prompt

The character sheet prompt that actually holds up in Seedance. 🎬

Higgsfield just open-sourced everything from their $2M AI feature film, and the biggest lesson was about character sheets: 3 panels (front, back, close portrait) — and TWO close-ups, one neutral and one smiling, or the model invents someone else's teeth.

I turned that into a master prompt for Nano Banana Pro on Higgsfield. Upload one photo of any person, paste the prompt, and you get a 4-panel sheet:

🧍 Front full body — head to shoes, hands visible
🔄 Back full body — same pose, same scale
😐 Close portrait, 3/4 view, neutral
😄 Close portrait, 3/4 view, genuine smile with real teeth

Same face, same outfit, same hair, same proportions in every panel. Neutral grey backdrop, even 5600K studio light, print-ready.

Works for male or female — the prompt adapts grooming and proportions automatically.

Tips:
* Use a clear front-facing photo with even lighting, no filters
* Set aspect ratio to 16:9 so four panels fit
* Pro trick from Higgsfield's own film: crop the head out of the two full-body panels so the model only copies the face from the close-ups
* Feed the finished sheet into Soul ID and your character stays locked across every scene

Prompt is below 👇

Create a professional character sheet based strictly on the attached reference photo. The sheet is one single image with four panels side by side on a plain neutral grey studio backdrop, separated by thin dark dividers, presented as a technical model turnaround with the exact realistic photographic style of the reference.

The subject is the same real person as in the attached photo, with the identical face, facial structure, skin tone, eye color, eyebrows, nose, lips, hairstyle, hair color, facial hair, glasses if present, age, height and body build. If the person is male, keep his male grooming and proportions exactly; if female, keep her female features, hairstyle and proportions exactly. Outfit: keep the same clothing as in the reference photo, with real fabric texture and natural creases; all clothing is opaque, closed and fully covering. If the reference only shows the upper body, complete the outfit with plain dark trousers and simple shoes that match the top.

Panel 1 — FRONT, full body: full-length front view, head to shoes, standing straight and neutral facing the camera, arms relaxed at the sides with both hands clearly visible and intact, feet visible, even framing with headroom and foot room, the whole outfit visible.

Panel 2 — BACK, full body: full-length back view, same standing pose, facing directly away from the camera, showing the back of the hair, the back of the outfit and the shoes from behind, same scale and alignment as panel 1.

Panel 3 — CLOSE PORTRAIT, neutral: large close-up portrait from the chest up in three-quarter view, face turned about 30 degrees, eyes looking toward the lens, calm neutral expression with lips closed, sharp detail on every facial feature, natural skin texture with pores, readable eye catchlights.

Panel 4 — CLOSE PORTRAIT, smiling: the same three-quarter close-up framing as panel 3, now with a genuine warm smile showing the person's real teeth, eyes slightly crinkled, same head angle and lighting.

Lighting: neutral soft even wrap-around studio light, near-shadowless, daylight-balanced 5600K, no color cast, identical across all four panels. Sharp realistic photograph, crisp and print-ready.

Locks: one identical person in all four panels — same identity, same outfit, same hair, same proportions; feet visible in both full-body panels; identical backdrop, scale and light in every panel; photoreal, no illustration or 3D look.

低饱和全身构图多分屏长发影棚光多面板布局影棚人物肖像