Character Consistency Across Images
Added 2026-10
Character Consistency Across Images — Chinese original prompt
为[角色描述]生成一组[数量]张图，场景不同但必须是同一人。请先写一段固定的外貌锚点：1）脸型、五官、发际线与发型细节；2）肤色与标志性特征（痣、疤痕、眼镜）；3）体型与身高感；4）固定服装元素；5）每张图都要重复的锚点句；6）允许随场景变化的变量。输出一段可直接复用的角色描述。
Site original
Chinese original
Example
Sample input
要给同一个角色生成六张不同场景的图，先帮我写一段固定外貌锚点
The output should include
- 锚点覆盖脸型、肤色与服装
- 给一句每张图都重复的锚点句
- 列出可以随场景变化的变量
Replace these placeholders 2
Swap the bracketed parts for your own details before sending.
[角色描述][数量]
How to use
把不变的特征写成固定段落，每张图原样粘贴，比形容词管用。
Try tweaking it
- Swap style words for measurable parameters: light, lens, material, composition.
- Ask for three variants of one prompt, then compare the differences.
What you should get
- A parameter sheet you can act on (light, lens, material, framing)
- Two or three alternatives with trade-offs
Common mistakes
- Stacking words like "premium, cinematic, 8K" — no two results match.
- Asking for a real brand or celebrity look and getting something close but unusable.
How to check the output
- Parameters are executable — ready to use in a tool or hand to a designer.
- It explains the trade-offs, not just the final look.
Where it works
- Good for: breaking vague "premium feel" into concrete light, material, lens and composition parameters.
- Watch out: for real brands, celebrities or third-party work, use equivalent descriptions to avoid infringement.