Meshy、Tripo 这类图生 3D 工具，几乎完全吃你喂进去的参考图。你的参考有多干净，出来的 mesh 就有多干净。
一张戏剧光、动态姿势、花背景的"好看插画"，到了 3D 生成器手里会变成一坨融化的几何体——腋下粘连、比例失真、背景也被建进网格。
所以这一步的目标不是画得漂亮，是画得"能被机器读懂"。投在这里的每一分钟，会在后面 8 步里成倍返还。
下面每一道工序，都是为了让最终参考满足这 5 条。先把规则记死，再动手。
手臂贴身躯，生成器分不清哪是手臂哪是身体，腋下会粘连。A-pose（手臂下垂 30–45°）通常比 T-pose 更稳，手腕也不容易和大腿穿模。
戏剧光、轮廓光会被直接烤进几何和贴图，进了 3D 改不掉。要的是均匀的影棚柔光。
纯白或纯灰，无地面、无投影、无道具。背景越干净，抠图和生成越准，否则背景会被一起建模。
平视、无大仰俯角。透视会让生成器误判比例——头大脚小、近大远小全乱套。
正 / 侧 / 背是同一个人：同配色、同装备、同比例。这是 AI 最难、也是本节核心战场——靠一致性模型 + 参考输入解决。
别一上来就抽卡。先用 ChatGPT 把"设定"写清楚：一句话概念、剪影特征、配色、标志性细节。文字越清楚，图越好出、越一致。直接套下面这条结构化 prompt：
You are a senior character designer for a stylized 3D action game.
Lock a production-ready character concept from my seed idea. Output:
1. One-paragraph concept (who they are, their world)
2. Silhouette features (what makes them readable as a black shape)
3. Color palette: 3-4 hex colors with roles (primary / secondary / accent)
4. 5 distinctive design details (gear, marks, materials)
5. A ready-to-paste FRONT-VIEW image prompt: full-body, A-pose,
plain white background, neutral studio lighting.
Seed idea: [你的一句话点子，例如 a wandering blue-cat spirit swordsman]
Keep it production-oriented: clean silhouette, separable parts,
no extreme anatomy that is hard to model.用结构化 prompt 生成正视图，一次多生几张，挑剪影最强、最符合设定的那张。结构永远是：主体 + 风格 + 姿势 + 取景 + 光 + 背景。
full-body character concept of [角色描述], front orthographic view, A-pose with arms held ~35° away from the body, legs slightly apart, character reference sheet, flat even studio lighting, soft shadows only, plain solid white background, clean readable design, [STYLE], symmetrical, entire figure visible head to feet, no cropping, high detail, sharp focus avoid: dramatic lighting, rim light, busy background, props, ground shadow, action pose, foreshortening, extreme perspective, multiple characters, text, watermark, cropped limbs
选中那张后，用图像编辑把它收干净，然后锁死，不再改设计。这张就是你的 master。
Using this image, keep the EXACT same character design, outfit and colors. Clean it up only: - remove the background to pure white - complete any cropped limbs so the full body (head to feet) is visible - flatten dramatic lighting into even neutral studio light - correct obvious proportion issues A-pose, front orthographic view. Do not redesign anything.
核心技巧：把 master 当参考图喂给 Nano Banana Pro（Gemini 3 Pro Image）——它的强项正是"角色身份锁定、跨图保持一致"，最多支持喂 14 张参考。让它只换视角、不改设计。
Use the provided reference image as the single source of truth for this
character. Generate the SAME character in a [side profile / back]
orthographic view. Keep identical outfit, colors, proportions, gear and
design details - this must read as the EXACT same character, just rotated.
A-pose, flat neutral studio lighting, plain white background,
character turnaround sheet style. Do not change or add any design element.Create a character turnaround / model sheet of THIS exact character showing front, side and back views in a row. All A-pose, identical design and colors, flat neutral lighting, plain white background, evenly aligned at the same height and scale.
武器、头盔、披风、大型配饰这类，单独抠出来、纯白底、正交。后面用更高质量单独生成 3D 再组装——比硬生一整个复杂角色干净太多。能拆就拆。
Isolate the [weapon / helmet / cape] from this character.
Show it ALONE on a plain white background, orthographic front and side
views, clean even studio lighting, no character, no background props,
product-shot style, high detail.规则照旧，prompt 里追加风格词即可。3D 生成最友好的一类。
semi-realistic game character, PBR-friendly, neutral expression, balanced anatomy注意：大眼、夸张比例会让 3D 生成更难，剪影要更克制。
anime style, cel-shaded, clean lineart, flat colors, readable silhouette比例可以夸张，但务必四肢分离、剪影清晰，否则圆乎乎一团会粘连。
chibi / stylized cartoon, exaggerated proportions, separated limbs, bold shapes▸ 三个都过 → 进入模块 02：3D 模型生成。