Image Create
Quick workflow
- 1. Confirm goal in one line (what image should achieve).
- Collect minimum constraints:
- subject
- style
- aspect ratio / resolution
- must-include elements
- must-avoid elements
- 3. Produce 3 prompt variants:
- safe/default
- bold/creative
- minimal/clean
- 4. Generate image(s) with the selected variant.
- Iterate with precise edits (lighting, framing, typography, colors, detail level).
Prompt template
Use this template and fill all placeholders:
CODEBLOCK0
Fast defaults
- - If user gives little detail, default to:
- style: photorealistic (real camera look, natural skin texture, true-to-life materials)
- lighting: soft natural light
- composition: centered medium shot
- lens/camera feel: 50mm, shallow depth of field, realistic dynamic range
- avoid: CGI/plastic look, over-smoothing, blur, watermark, extra limbs, unreadable text, distorted faces
- - If text appears in image, keep wording short and ask for exact spelling.
Realistic style tuning
Use these additions whenever the user asks for realistic output:
- - Add quality cues:
photorealistic, natural textures, physically plausible lighting, realistic shadows. - Add camera cues:
shot on full-frame camera, 50mm lens (or 85mm for portraits). - Keep color grading subtle; avoid neon oversaturation unless requested.
- Prefer environmental detail that supports realism (imperfections, material grain, natural reflections).
Iteration rules
- - Change one major variable per iteration when debugging quality.
- Keep successful parts of previous prompt; only patch weak parts.
- For brand visuals, preserve palette and typography constraints across iterations.
- For thumbnails/posters, prioritize readability and high contrast.
Output format
When helping with generation requests, return:
- 1. Primary prompt
- 2 alternatives
- Negative prompt
- Suggested settings (aspect ratio, style strength, number of outputs)
- One-line revision plan for next iteration
图像创建
快速工作流程
- 1. 用一句话确认目标(图像应实现什么效果)。
- 收集最低约束条件:
- 主体
- 风格
- 宽高比/分辨率
- 必须包含的元素
- 必须避免的元素
- 3. 生成3个提示词变体:
- 安全/默认
- 大胆/创意
- 极简/干净
- 4. 使用选定的变体生成图像。
- 通过精确调整进行迭代(光照、构图、排版、色彩、细节程度)。
提示词模板
使用此模板并填写所有占位符:
text
[图像类型],以[主要主体]为特色,采用[风格],
构图:[相机/取景],光照:[光照],
背景:[背景],调色板:[色彩],
氛围:[氛围],质量:高细节,边缘清晰,
输出:[宽高比或尺寸]。
避免:[负面约束]
快速默认设置
- 风格:照片级真实感(真实相机效果,自然皮肤纹理,逼真的材质)
- 光照:柔和的自然光
- 构图:居中中景
- 镜头/相机感:50mm,浅景深,真实的动态范围
- 避免:CGI/塑料感,过度平滑,模糊,水印,多余肢体,不可读文字,扭曲的面部
- - 如果图像中出现文字,保持措辞简短并要求准确拼写。
真实风格调整
每当用户要求真实输出时,使用以下补充:
- - 添加质量提示:照片级真实感,自然纹理,物理上合理的光照,真实的阴影。
- 添加相机提示:使用全画幅相机拍摄,50mm镜头(人像用85mm)。
- 保持色彩分级微妙;除非要求,否则避免霓虹色过度饱和。
- 优先选择支持真实感的环境细节(瑕疵、材质纹理、自然反射)。
迭代规则
- - 调试质量时,每次迭代只改变一个主要变量。
- 保留上一个提示词中成功的部分;仅修补薄弱部分。
- 对于品牌视觉,在迭代中保持调色板和排版约束。
- 对于缩略图/海报,优先考虑可读性和高对比度。
输出格式
在帮助处理生成请求时,返回:
- 1. 主要提示词
- 2个备选方案
- 负面提示词
- 建议设置(宽高比、风格强度、输出数量)
- 一行修订计划,用于下一次迭代