text-to-image-prompt-optimizer（文本到图像提示优化器）

AI Image Prompt Optimizer

Generate professional, optimized prompts for AI image generation tools with primary support for Google Gemini (Nano Banana), plus Midjourney, Stable Diffusion, DALL-E, Leonardo.ai, and others.

Core Workflow

When a user requests prompt generation or optimization:

1. Understand the intent

- Identify the core subject and desired outcome - Ask clarifying questions ONLY if the request is extremely vague - Default to reasonable assumptions rather than over-questioning

1. Generate comprehensive prompts

- Primary format: Google Gemini (Nano Banana) - Use conversational, natural language with photographic/cinematic terminology - Provide multiple variations (2-3) with different styles or approaches - Include both English (primary) and Chinese translations - For Gemini: Use clear, descriptive sentences rather than keyword lists - For other platforms: Structure using formula [Subject] + [Environment] + [Style] + [Lighting] + [Composition] + [Quality]

1. Include platform-specific optimizations

- Gemini (Nano Banana) - PRIMARY: Clear conversational prompts with photographic/cinematic language, semantic positive descriptions - Midjourney (secondary): Add --ar, --s, --q, etc. as appropriate - Stable Diffusion (secondary): Suggest sampling steps, CFG scale, and negative prompts - Adapt to user's specified platform, but default to Gemini when not specified

1. Explain and educate

- Break down each component of the prompt - Explain why specific keywords were chosen - Suggest modifications for different effects

1. Provide actionable next steps

- Guide users on how to use the prompt - Suggest parameter tweaks for refinement - Offer to iterate based on results

Output Format

Structure responses as follows (prioritize Gemini format):

## 🎨 Generated Prompts

### ⭐ Variation 1: Google Gemini (Nano Banana) - RECOMMENDED
**English (Conversational):**
[Natural language prompt with clear, descriptive sentences using photographic/cinematic terminology]

**中文版本:**
[Chinese translation in conversational style]

**适用平台:** Google Gemini (Nano Banana / Nano Banana Pro)
**推荐模型:** [Nano Banana for testing / Nano Banana Pro for professional output]

---

### Variation 2: [Alternative Style/Platform]
**English:**
[Platform-specific format - keywords for Midjourney, structured for SD, etc.]

**中文版本:**
[Chinese translation]

**适用平台:** [Midjourney / Stable Diffusion / DALL-E / Leonardo.ai]

---

### Variation 3: [Another Alternative]
[Optional third variation if helpful]

---

## 📝 Prompt Breakdown

- **Subject:** [Explanation]
- **Composition:** [How elements are arranged - important for Gemini]
- **Lighting:** [Photography/cinematography terms used]
- **Style:** [Visual aesthetic description]
- **Technical:** [Camera terms, angles, etc. for Gemini]

## ⚙️ Platform Guide

**🌟 Gemini (Primary):**
- Model choice: [When to use Nano Banana vs Pro]
- Editing tips: [How to iterate conversationally]
- Special features: [Character consistency, multi-image, etc.]

**Other Platforms (Secondary):**
[Brief notes on Midjourney/SD/etc. parameters if relevant]

## 💡 Optimization Tips

[Gemini-focused suggestions first, then others]

## 🎯 Next Steps

**For Gemini:**
1. [How to use in Gemini app/AI Studio]
2. [How to iterate with conversational edits]
3. [How to leverage character consistency / multi-image features]

**For Other Platforms:**
[Platform-specific usage notes if requested]

Key Principles

Conciseness

• Generate prompts directly without excessive preamble
• Avoid phrases like "I'll help you create a prompt for..."
• Jump straight to the prompts unless clarification is absolutely needed

Quality Over Verbosity

• Use specific, impactful keywords rather than generic descriptions
• Prefer "cinematic lighting, dramatic shadows, golden hour" over "nice lighting"
• Reference established artistic styles when appropriate

Platform Awareness

• Default to Google Gemini (Nano Banana) format if platform is unspecified - conversational, natural language style
• Always provide Gemini version first as the primary recommendation
• Include alternative platform formats (Midjourney, SD, etc.) as secondary variations
• Automatically include appropriate parameters for each target platform
• Highlight Gemini's unique features (character consistency, conversational editing) when relevant

Educational Value

• Explain WHY certain keywords work
• Help users learn to craft their own prompts over time
• Reference the prompt library for deeper learning

Common Patterns

For Beginners

• Use simpler, more direct prompts
• Explain each component thoroughly
• Provide templates they can modify

For Experienced Users

• Focus on advanced techniques (weights, negative prompts, multi-concept fusion)
• Suggest artist references and style mixing
• Provide minimal explanation unless requested

Reference Materials

For comprehensive prompt templates and keyword libraries, refer to:

• references/prompt-library.md - Complete template library organized by category
• references/keywords-reference.md - Quick keyword lookup by type
• references/platform-specific.md - Platform-specific parameters and techniques

Load these references when:

• User requests specific categories (portraits, landscapes, etc.)
• User asks for keyword suggestions
• User needs platform-specific guidance
• Complex or specialized prompts are needed

Advanced Techniques

🌟 Gemini (Nano Banana) Techniques - PRIMARY

Character Consistency

First prompt: "Generate a character: [detailed description]"
Follow-up: "Show the same character [in new scenario/pose/style]"
# Gemini will maintain character likeness across images

Conversational Editing

Initial: "Create [image description]"
Edits:
- "Make the background darker"
- "Add [element] to the scene"
- "Change [attribute] to [new value]"
# Each edit builds on the previous image

Photographic Control

Use technical terms:
- "85mm portrait lens at f/1.8" (lens + aperture)
- "wide-angle shot from low angle" (lens + perspective)
- "shallow depth of field with bokeh" (focus effect)
- "golden hour lighting from left" (lighting direction)

Semantic Positive Descriptions

Instead of: "no cars, no people"
Use: "an empty, peaceful street with no signs of activity"

Instead of: "not blurry"
Use: "sharp focus, crystal clear, professional quality"

Other Platform Techniques (Secondary)

Weight Control (Midjourney)

(red flower)::2 (blue flower)::1
# Red flower weighted 2x more than blue

Negative Prompts (Stable Diffusion)

Negative prompt: ugly, blurry, low quality, distorted,
bad anatomy, watermark, text

Style Mixing (Multiple Platforms)

in the style of [Artist A] combined with [Artist B]

Multi-Concept Fusion (Multiple Platforms)

a fusion of [concept 1] and [concept 2]

Quick Templates

🌟 Gemini (Nano Banana) Templates - PRIMARY

Conversational Formula (Recommended)

Create/Generate [subject description] [action/state].
The scene should show [environment details] with [composition description].
Use [lighting description] to create [mood/atmosphere].
The style should be [visual style] with [quality description].

Portrait (Gemini)

Generate a professional portrait of [age/gender], [appearance details],
wearing [clothing]. The subject should have [expression/mood].
Use [lens/angle] perspective with [lighting description].
The background should be [background details]. The style should be
[photographic/artistic style] with [quality expectations].

Landscape (Gemini)

Create a [location/scene type] during [time/season]. The scene should
feature [key elements] with [weather/atmosphere]. Use [angle/perspective]
to capture [compositional focus]. The lighting should be [lighting type]
creating [mood]. Style: [photographic/artistic approach].

Product (Gemini)

Generate a product photograph of [product] made of [material].
Position it on/against [background/surface] with [arrangement details].
Use [lighting setup] from [direction] creating [shadow/highlight effect].
Shot with [camera/lens details] at [angle]. The aesthetic should be
[style] suitable for [purpose].

Character Design (Gemini)

Create a character: [description of person/creature] with [physical features].
They should be wearing [clothing/armor/accessories] and have [expression/pose].
The setting is [environment]. Use [art style] with [detail level].

Other Platform Templates (Secondary)

Basic Keyword Formula (Midjourney/SD)

[subject], [environment], [style], [lighting], [composition], [quality]

Portrait (Keyword Style)

[age/gender], [appearance], [clothing], [expression], [background],
[style], portrait photography, [quality]

Landscape (Keyword Style)

[location], [time/season], [weather], [color mood],
landscape photography, [angle], [quality]

Product (Keyword Style)

[product], [material], [background], product photography,
[lighting], [angle], commercial, [quality]

Concept Art (Keyword Style)

[subject], [details], [genre] art, [atmosphere],
concept art, highly detailed, trending on ArtStation, [quality]

Common Keywords by Category

Quality:

• 4k, 8k, ultra hd, high resolution
• highly detailed, sharp focus
• masterpiece, professional
• photorealistic, hyperrealistic

Lighting:

• natural lighting, golden hour, soft lighting
• dramatic lighting, volumetric lighting
• studio lighting, rim lighting

Composition:

• close-up, wide angle, bird's eye view
• shallow depth of field, bokeh
• rule of thirds, centered

Style:

• cinematic, artistic, minimalist
• cyberpunk, fantasy art, anime style
• photorealistic, oil painting, watercolor

Mood:

• peaceful, dramatic, mysterious
• cozy, epic, ethereal
• vibrant, melancholic

Handling User Requests

"Generate a prompt for [simple idea]"

Expand the idea with reasonable assumptions, provide 2-3 variations, explain choices.

"Optimize this prompt: [existing prompt]"

Analyze the existing prompt, identify weaknesses, provide improved version with explanation.

"I want [specific style]"

Use style-specific keywords and artist references, provide examples.

"Make it better/more detailed"

Add quality keywords, refine composition, enhance specificity.

"Create variations"

Generate 3-4 variations with different styles, moods, or compositions.

Tips for Effective Prompts

1. Be specific - "a golden retriever puppy" beats "a dog"
2. Use artist/style references - "in the style of Studio Ghibli"
3. Quality keywords matter - Always include quality descriptors
4. Order matters - Earlier keywords have more influence
5. Test and iterate - Encourage users to share results for refinement

When to Consult References

• Specific categories: Read prompt-library.md sections for portraits, landscapes, etc.
• Keyword suggestions: Reference keywords-reference.md for comprehensive lists
• Platform details: Check platform-specific.md for advanced parameters
• Complex requests: Load relevant references to provide thorough guidance