在AI創(chuàng)作的世界里，提示詞是開啟創(chuàng)意大門的鑰匙。本文將帶你解鎖如何用Gemini 2.5打造高質(zhì)量的4o提示詞，輕松生成獨具風格的AI作品，無論是彩繪風還是科普卡，都能信手拈來，快來一探究竟吧！

最近一直在用4o抽各種好玩的卡，但每個提示詞都要單獨寫，于是我就在琢磨能不能搞個通用提示詞和穩(wěn)定的風格出來。

剛好最近刷到了一些彩繪風格的圖片，我覺得挺不錯的，我就開始研究彩繪海報要怎么搞；于是我拉上我的多模態(tài)好搭子Gemini2.5來寫4o的提示詞，搞了一晚上搗騰出來一個AI彩繪風提示詞出來。

這個過程還挺有意思的，用AI來生成AI的提示詞，人類負責提需求和標注。

那讓我們跟隨一個AI漫畫科普提示詞來看一看如何用Gemini2.5來寫提示詞吧：

1.打開Gemini開發(fā)者模式，選擇2.5版本：

https://aistudio.google.com/app/prompts/new_chat

2.輸入自己的需求，最好帶上參考樣例，這樣可以讓AI更加聚焦。

我這里用的是 @吳熳 老師做的英語卡片，看到了她的卡片，我感覺我可以搞點AI科普卡片出來。

繼續(xù)和AI溝通，告訴它我們到底要做一件什么樣的事情。

3.根據(jù)AI產(chǎn)出的第一版提示詞，扔到4o里去測試，看看效果怎么樣

4.拿到4o的結(jié)果后，人要給到足夠的建議，給Gemini 2.5進行標注，方便Gemini了解如何去優(yōu)化提示詞

5.反復標注改提示詞，直到生成一個滿意的為止。

需求越復雜，標注次數(shù)越多，我的AI繪畫卡標注了十幾輪次的對話。

6.標注完了讓AI繼續(xù)產(chǎn)出多個其它內(nèi)容的提示詞，我們?nèi)y試效果怎么樣，矯正提示詞的穩(wěn)定性。

接下來可以選擇繼續(xù)讓AI產(chǎn)出提示詞，然后一個卡片一個卡片生成。

也可以再萃取一個通用提示詞出來，萃取通用提示詞我們需要告訴AI明確的需求和結(jié)構(gòu)要求；如果有特定的結(jié)構(gòu)比如我喜歡json，那我會給到參考json。

最后難道AI給你的json提示詞，繼續(xù)去4o做測試。

重復標注+測試流程若干次，得到一個穩(wěn)定性高的提示詞。

以下是一些使用Gemini 2.5的技巧：

1.記得點save保存，要么刷新網(wǎng)頁所有內(nèi)容都沒了。

Gemini的開發(fā)者模式是沒有自動保存的，如果你跟AI嘮嗑了半天沒保存不小心關(guān)了網(wǎng)頁，那么你只能從頭開始了。

2.覺得Gemini本次回答的不好，可以點這個按鈕，讓它重新生成

最后是兩個提示詞，大家需要的自取，全部開源：

– AI彩繪提示詞：

## 核心指令：基于以下內(nèi)容，使用定義的藝術(shù)風格配置文件創(chuàng)作插畫 (中文內(nèi)容)

**【用戶提供的內(nèi)容】**：

[ ?這里填寫內(nèi)容? ]

—

**【藝術(shù)風格配置文件】(Art Style Profile):**

{
“art_style_profile”: {

“style_name”: “Expressive Hand-Drawn Concept Illustration (Rich Colored Pencil)”,

“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, 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 (簡體中文).** 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

– AI科普卡提示詞：

## 核心指令：為用戶提供的AI術(shù)語中文名稱，自動創(chuàng)作并生成一張解釋卡片插畫。

**第一步：內(nèi)容創(chuàng)作（由AI完成）**

* **輸入：** 用戶僅提供AI術(shù)語的中文名稱： **[ ?強化學習? ]**

* **任務(wù)：**

1. **確定英文名稱：** 查找或生成該術(shù)語對應(yīng)的常用英文名稱。

2. **撰寫中文定義：** 創(chuàng)作一句簡潔、清晰、易于理解的中文定義，解釋該術(shù)語的核心含義。定義應(yīng)適合科普性質(zhì)的卡片。

3. **構(gòu)思插畫場景：**

* 思考如何用一個具體的、視覺化的場景或比喻來解釋這個術(shù)語。

* 場景應(yīng)易于理解，能直觀傳達術(shù)語的關(guān)鍵概念。

* 描述這個場景，包括必要的元素（人物、物品、動作、符號等）及其互動方式。**這個描述將作為下面圖像生成的主要依據(jù)。**

**第二步：圖像生成（基于第一步創(chuàng)作的內(nèi)容，使用定義的藝術(shù)風格）**

* **使用以下【藝術(shù)風格配置文件】和第一步生成的內(nèi)容（英文名、定義、插畫場景描述）來創(chuàng)作最終的卡片圖像。**

—

**【藝術(shù)風格配置文件】(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