在AI创作的天下里杨超越 ai换脸，教唆词是开启创意大门的钥匙。本文将带你解锁怎样用Gemini 2.5打造高质地的4o教唆词，轻便生成独具作风的AI作品，不管是彩绘风也曾科普卡，齐能信手拈来，快来一探究竟吧！
图片
最近一直在用4o抽各式好玩的卡，但每个教唆词齐要单独写，于是我就在沟通能不行搞个通用教唆词和康健的作风出来。
刚好最近刷到了一些彩绘作风的图片，我认为挺可以的，我就开动讨论彩绘海报要怎样搞；于是我拉上我的多模态好搭子Gemini2.5来写4o的教唆词，搞了一晚上捣腾出来一个AI彩绘风教唆词出来。
图片
这个过程还挺迥殊义的，用AI来生成AI的教唆词，东谈主类厚爱提需乞降标注。
图片
那让咱们随从一个AI漫画科普教唆词来看一看怎样用Gemini2.5来写教唆词吧：
1.掀开Gemini建造者格式，遴荐2.5版块：
https://aistudio.google.com/app/prompts/new_chat
图片
2.输入我方的需求，最佳带上参考样例，这么可以让AI愈加聚焦。
我这里用的是 @吴熳 栽种作念的英语卡片，看到了她的卡片，我嗅觉我可以搞点AI科普卡片出来。
图片
连接和AI交流，告诉它咱们到底要作念一件什么样的事情。
图片
图片
3.证据AI产出的初版教唆词，扔到4o里去测试，望望边界怎样样
图片
4.拿到4o的边界后，东谈主要给到有余的提议，给Gemini 2.5进行标注，浅易Gemini了解怎样去优化教唆词
图片
5.反复标注改教唆词，直到生成一个赋闲的为止。
需求越复杂，标注次数越多，我的AI绘图卡标注了十几轮次的对话。
图片
6.标注收场让AI连接产出多个其它实践的教唆词，咱们去测试边界怎样样，改良教唆词的康健性。
图片
图片
接下来可以遴荐连接让AI产出教唆词，然后一个卡片一个卡片生成。
也可以再萃取一个通用教唆词出来，萃取通用教唆词咱们需要告诉AI明确的需乞降结构条目；若是有特定的结构比如我心爱json，那我会给到参考json。
图片
终末难谈AI给你的json教唆词，连接去4o作念测试。
图片
重叠标注+测试进程多少次，获取一个康健性高的教唆词。
以下是一些使用Gemini 2.5的工夫：
1.铭记点save保存，要么刷新网页总共实践齐没了。
Gemini的建造者格式是莫得自动保存的，若是你跟AI唠嗑了半天没保存不预防关了网页，那么你只可重新开动了。
图片
2.认为Gemini本次回应的不好，可以点这个按钮，让它再行生成
图片
终末是两个教唆词，大家需要的自取，一起开源：
– AI彩绘教唆词：
## 中枢指示：基于以下实践，使用界说的艺术作风建立文献创作插画 (汉文实践)
**【用户提供的实践】**：
[ 这里填写实践 ]—
**【艺术作风建立文献】(Art Style Profile):**
{ “art_style_profile”: {
“style_name”: “Expressive Hand-Drawn Concept Illustration (Rich Colored Pencil)”，
推特 男同“visual_elements”: {
“shape_language”: “Organic forms， flowing lines， visible hand-drawn strokes and imperfections， avoids rigid geometry and perfect symmetry.”，
“colors”: {
“primary_palette”: [“Vibrant， warm， and energetic (e.g.， oranges， yellows， warm reds， lively blues/greens) OR adaptable palette based on the concept’s mood.”， “Natural blending and layering typical of colored pencils.”]，
“accent_colors”: [“Used dynamically for emphasis杨超越 ai换脸， often contrasting or complementary.”， “Potential use of symbolic colors like gold/orange if relevant to concept.”]，
“shading”: “Achieved through layered colored pencil strokes (hatching， cross-hatching， scumbling). Visible stroke direction and texture are KEY. Avoids smooth digital gradients.”
}，
“lighting”: {
“type”: “Atmospheric and expressive， enhances mood and form rather than strict realism.”，
“source_direction”: “Variable， dictated by compositional needs and desired mood.”，
“shadow_style”: “Textured shadows built up with pencil strokes， edges can be soft or defined based on stroke density， reflecting the medium.”
}，
“materials”: {
“surface_texture”: “**MUST HAVE prominent paper or canvas texture visible throughout.** Pigment texture from colored pencils should be apparent.”，
“reflectivity”: “Very low， matte finish typical of physical media on paper.”
}，
“composition”: {
“object_presentation”: “Dynamic， organic arrangement. Elements should feel interconnected， flowing， or actively interacting， centered around a core visual metaphor. **Strictly avoid grid-like or list-like layouts.** **Visual concept takes priority over displaying text.**”，
“perspective”: “Often employs flattened perspective or slight expressive distortion， avoids rigid vanishing points unless conceptually necessary.”，
“background”: “**Visually integrated and conceptually supportive.** Must feature visible hand-drawn texture and strokes. However， the specific execution (e.g.， pattern， dominant colors， structure， density of strokes) should **vary significantly based on the user’s content theme and mood**， offering visual distinction between different illustrations. **Avoid repetitive default background treatments (like generic swirls for every image); let the background subtly enhance the specific subject and atmosphere.**”
}，
“typography”: {
“font_style”: “**Highly legible Hand-drawn artistic Chinese characters (简体汉文).** Clarity is paramount. Style should match the illustration’s energy but prioritize readability. Avoid overly complex or illegible styles.”，
“text_placement”: “**Include a short， descriptive phrase or tagline (perhaps 1-2 short lines， roughly 8-20 characters total， flexible) derived from the user’s content that captures the core purpose， benefit， or feeling.** This text should be clearly readable， artistically integrated as a significant visual element， adding descriptive value beyond just a title. It must enhance understanding without overwhelming the visuals. **Aim for meaningful description， not just keywords.**”，
“color”: “Harmonizes with the overall color palette， ensuring **strong contrast** for readability.”
}，
“rendering_style”: {
“technique”: “**Simulated Rich Colored Pencil on Textured Paper.** Focus on capturing the tactile quality.”，
“detail_level”: “Medium to high detail in texture and stroke work， while forms remain expressive and potentially simplified. Focus on 'soul’ over photorealism.”
} }，
“purpose”: “To generate a conceptually driven， emotionally resonant illustration that leverages the unique aesthetic and tactile qualities of hand-drawn colored pencil art to interpret the user’s input， **communicating primarily through visuals， thoughtfully enhanced by a meaningful， descriptive， yet concise hand-drawn text phrase or tagline that is seamlessly integrated.**”，
“negative_constraints”: [
“No photographic elements.”，
“No smooth， vector-like gradients or fills.”，
“No perfectly straight lines or geometric precision (unless part of the concept).”，
“No standard computer fonts.”，
“No non-Chinese text.”，
“Avoid overly clean， sterile digital look.”，
“Avoid literal， uninspired interpretations of the input.”，
“**Avoid lengthy sentences， multiple paragraphs， or purely informational text blocks.**”，
“**Avoid text that is difficult to read due to style， size， or placement.**”，
“**Avoid layouts where text visually dominates the artwork.**”，
“**Avoid highly repetitive background patterns across different images unless conceptually justified.**”
] }
} –ar 9:16杨超越 ai换脸
– AI科普卡教唆词：
## 中枢指示：为用户提供的AI术语汉文称号，自动创作并生成一张证据卡片插画。
**第一步：实践创作（由AI完成）**
* **输入：** 用户仅提供AI术语的汉文称号： **[ 强化学习 ]*** **任务：**
1. **笃定英文称号：** 查找或生成该术语对应的常用英文称号。
2. **撰写汉文界说：** 创作一句简洁、了了、易于相识的汉文界说，证据该术语的中枢含义。界说应符合科普性质的卡片。
3. **构念念插画场景：**
* 念念考怎样用一个具体的、视觉化的场景或譬如来证据这个术语。
* 场景应易于相识，能直不雅传达术语的要道见识。
* 描述这个场景，包括必要的元素（东谈主物、物品、动作、记号等）过甚互动面貌。**这个描述将动作底下图像生成的主要依据。**
**第二步：图像生成（基于第一步创作的实践，使用界说的艺术作风）**
* **使用以下【艺术作风建立文献】和第一步生成的实践（英文名、界说、插画场景描述）来创作最终的卡片图像。**
—
**【艺术作风建立文献】(Art Style Profile):**
{ “art_style_profile”: {
“style_name”: “Warm FriendlyWatercolor Educational Card”，
“visual_elements”: {
“shape_language”: “Soft edges， friendly and approachable forms， simple and clear shapes suitable for easy understanding.”，
“colors”: {
“primary_palette”: [“Warm and inviting colors (e.g.， soft oranges， yellows， gentle blues， light greens). Palette should feel cohesive and pleasant.”]，
“accent_colors”: [“Used subtly for emphasis where needed， maintaining overall harmony.”]，
“shading”: “Achieved through soft watercolor washes and layering. Avoid harsh lines or complex gradients. Texture should suggest watercolor on paper.”
}，
“lighting”: {
“type”: “Simple， clear， and even lighting. Focus is on clarity， not dramatic effect.”，
“source_direction”: “Generally frontal or ambient.”，
“shadow_style”: “Soft， subtle shadows using watercolor washes.”
}，
“materials”: {
“surface_texture”: “**MUST HAVE a visible， consistent textured paper background throughout the card.** Illustrations should have a gentle watercolor texture.”，
“reflectivity”: “Low， matte finish typical of watercolor on paper.”
}，
“composition”: {
“object_presentation”: “**Central illustration clearly depicting the concept described in the AI-GENERATED 'illustration_concept’ from Step 1.** Focus on the interaction or process being explained. Layout should be clean and uncluttered.”，
“perspective”: “Simple， often flattened 2D perspective for clarity.”，
“background”: “Subtle， unobtrusive background， often a light color wash or the textured paper itself， ensuring focus remains on the central illustration and text.”
}，
“typography”: {
“font_style”: “**Legible， friendly font for all text (Chinese and English).** Simplicity and readability are paramount. Avoid overly decorative fonts.”，
“text_placement”: [
“**Top:** Large， bold title containing the AI Term (Chinese first， AI-GENERATED English name below or beside in parentheses).”，
“**Below Title:** Smaller font size for the AI-GENERATED definition.”，
“**NO other text elements** unless part of the AI-GENERATED illustration concept itself.”
]，
“color”: “Dark， highly readable color (e.g.， dark brown， black， or a dark tone harmonious with the palette) ensuring strong contrast against the background.”
}，
“rendering_style”: {
“technique”: “**Simulated Warm Watercoloron Textured Paper.** Focus on a gentle， friendly， hand-drawn feel.”，
“detail_level”: “Medium detail， focusing on clarity and appeal rather than complexity. Forms should be easily recognizable.”
} }，
“purpose”: “To automatically generate a clear， friendly， and visually appealing educational flashcard for a given AI term (provided in Chinese). The process involves AI first creating the definition and illustration concept， then generating the image in a warm watercolor style.”，
“negative_constraints”: [
“No photographic elements.”，
“No vector art or smooth digital gradients.”，
“Avoid overly complex or abstract illustrations.”，
“Ensure generated definition is simple and accurate.”，
“Ensure generated illustration concept is relevant and explanatory.”，
“Avoid text that is hard to read.”，
“Strictly avoid adding extra text elements beyond title， definition， and those integral to the illustration concept.”，
“Avoid cluttered layouts.”
] }
} –ar 9:16
本站仅提供存储职业，总共实践均由用户发布，如发现存害或侵权实践，请点击举报。