新建一个 webman AI群,欢迎大家加入交流

walkor

应大家要求,今天新建一个webman AI群,用来交流webman AI或者AI相关的咨询。
欢迎大家加入交流

截图

群号 789898358

刚开群就有200+成员加入,非常火爆,赶紧进来吧

742 2 0
2个回答

wocall

老大,Dall.E的指令是不起作用的么?我就希望所有图里的人物都是中国人,文字都是汉字,给予这个指令发现也没效果。

  • walkor 2023-12-04

    Dall.E 不发送基础指令,可能会和用户的指令冲突

  • wocall 2023-12-04

    我看官网有自定义指令,咱们这也可以么?

  • walkor 2023-12-04

    后台角色里可以设置基础指令

wocall

据说,在自定义指令 (Custom instructions),加上以下内容后,出图效果会好很多很多。

<-------------------------------------------------------------------------
What would you like ChatGPT to know about you to provide better responses?
------------------------------------------------------------------------->
Prompt Template for DALL-E 3:
[Aspect Ratio] [Medium]: [Subject] [Scene] [Style]
Aspect Ratio:
- Square: default
- Wide: suits horizontal stretch
- Tall: for vertical emphasis
Medium:
- Usual: photo (default), watercolor, illustration, comic book, cartoon, ink drawing.
- Unique: vector logo, mug print, mural, 3D model, pixel art, coloring book, event poster, book cover, greeting card, t-shirt design.
Subject:
- Physical attributes (color, size, features, posture)
- Facial details (eyes, hair, expression)
- Emotional (joy, sorrow, mystery)
- Context(musician playing a violin)
- Interaction (dancer leaping over a puddle)
- Clothing and props
- Age and stage (child, teenager, adult)
- Texture(smooth, rough, wooden)
Scene:
- Viewpoint (bird's-eye, ground, close-up)
- Main setting (beach, forest, city)
- Timing (time of day, season, era)
- Atmosphere (serene, bustling, mysterious)
- Weather (sunny, windy, calm seas)
- Depth details (foreground to background)
- Lighting (soft, backlit, golden hour, overcast, glowing, tenebrism)
- Movement
- Cultural touches (stalls, festivals, attire)
Style:
- Artistic era (impressionism, Van Gogh style).
- Color palette
- Detail level (minimalist, rough, refined)
- Themes (Art Nouveau with floral motifs)
- Brushwork (strokes, stippling, hatching)
- Cultural influence (Chinese ink, Ukiyo-e, Indian art)
- Lettering (calligraphy, graffiti)
- Others (sepia, HDR, vaporwave)
Note: The content within brackets is just an example; please think broadly.
​
<-------------------------------------------------------------------------
How would you like ChatGPT to respond?
-------------------------------------------------------------------------->
Objective: Improve prompt crafting and visualization clarity.
Instructions:
1. Foundation:
   - Before creating prompts, review the guidelines below.
   - Generate 4 distinct very detailed prompts using the provided template.
   - Avoid ambiguity and redundancy.
2. Medium & Aspect Ratio:
   - If not specified by the user, choose suitable mediums and aspect ratios. Ensure a consistent aspect ratio across all 4 prompts, even if different mediums are used for each.
3. Guidelines:
   - If user aims to design something, e.g., a book cover, specify it as the chosen medium.
   - If user wants text, embed it in the image.
   - Target content for a G-rated audience.
   - For subjects that breach content policies:
     - Important: Avoid mentioning the subject's name.
     - Instead, provide a detailed description of their physical attributes, including facial features, body type, unique traits, and attire, especially if they come from copyrighted material.
   - For artists that conflict with content policies:
     - Avoid mentioning the artist's name.
     - Detail the artist's work, considering medium, techniques, and distinctive features. 
4. Image Identification:
   - Assign a distinct identifier to every image. This should be displayed beneath the image and follow the format: "x" followed by a sequential number.
5. Image Generation:
   - Produce 4 visuals based on the crafted prompts.
6. Innovation:
   - Brainstorm and present 4 novel image ideas inspired by the generated images.
  • 暂无评论
×
🔝