nano-banana

horuz-ai 的头像作者 horuz-ai

使用Google的Nano Banana(Gemini 2.5 Flash Image)和Nano Banana Pro(Gemini 3 Pro Image)API进行AI图像生成和编辑。当用户想要使用AI生成、编辑或合成图像时使用此技能。触发条件包括从文本描述创建图像的请求、编辑现有图像、从照片添加/移除元素、应用风格转换、保持图像间角色一致性、生成带有文本覆盖的图像(标志、海报、信息图)或创建多图像合成。当用户提及"Nano Banana"、"Gemini图像"或想要AI生成的视觉效果时也使用

📖 中文导读:使用Google的Nano Banana(Gemini 2.5 Flash Image)和Nano Banana Pro(Gemini 3 Pro Image)API进行AI图像生成和编辑。当用户想要使用AI生成、编辑或合成图像时使用此技能。触发条件包括从文本描述创建图像的请求、编辑现有图像、从照片添加/移除元素、应用风格转换、保持图像间角色一致性、生成带有文本覆盖的图像(标志、海报、信息图)或创建多图像合成。当用户提及"Nano Banana"、"Gemini图像"或想要AI生成的视觉效果时也使用 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: nano-banana description: "AI image generation using Nano Banana PRO (Gemini 3 Pro Image) and Nano Banana (Gemini 2.5 Flash Image). Use this skill when: (1) Generating images from text prompts, (2) Editing existing images, (3) Creating professional visual assets like infographics, logos, product shots, stickers, (4) Working with character consistency across multiple images, (5) Creating images with accurate text rendering, (6) Any task requiring AI-generated visuals. Triggers on: 'generate image', 'create image', 'make a picture', 'design a logo', 'create infographic', 'AI image', 'nano banana', or any image generation request."

Nano Banana PRO Image Generation

Generate professional AI images using Google's Nano Banana models via the Gemini API.

Prerequisites

  • API key must be set as GEMINI_API_KEY environment variable
  • Uses curl for all API calls (no SDK required)

Model Selection

Model Identifier Best For
Nano Banana PRO gemini-3-pro-image-preview Professional assets, text rendering, infographics, 4K output, complex multi-turn editing
Nano Banana gemini-2.5-flash-image Fast generation, simple edits, lower cost

Default to PRO for quality work. Use Flash for rapid iterations or simple tasks.

CRITICAL: Prompt Engineering First

BEFORE calling the API, always craft an effective prompt. Read references/prompting-guide.md for comprehensive prompting strategies. Key principles:

The Golden Rules

  1. Describe scenes, don't list keywords - Write narrative descriptions, not tag soup
  2. Use natural language - Full sentences with proper grammar
  3. Be specific - Define subject, setting, lighting, mood, materials
  4. Provide context - The "why" helps the model make better artistic decisions
  5. Edit, don't re-roll - If 80% correct, ask for specific changes

The ICS Framework (Quick Reference)

For any image, specify:

  • Image type: What kind of visual (photo, infographic, logo, sticker, etc.)
  • Content: Specific elements, data, or information to include
  • Style: Visual style, color palette, artistic approach

API Reference

Text-to-Image Generation

curl -s -X POST \
  "https://generativelanguage.googleapis.com/v1beta/models/gemini-3-pro-image-preview:generateContent" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{
      "parts": [{"text": "YOUR_PROMPT_HERE"}]
    }],
    "generationConfig": {
      "responseModalities": ["TEXT", "IMAGE"],
      "imageConfig": {
        "aspectRatio": "16:9",
        "imageSize": "2K"
      }
    }
  }'

Image Editing (with input image)

curl -s -X POST \
  "https://generativelanguage.googleapis.com/v1beta/models/gemini-3-pro-image-preview:generateContent" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{
      "parts": [
        {"text": "YOUR_EDIT_INSTRUCTION"},
        {"inline_data": {"mime_type": "image/png", "data": "BASE64_IMAGE_DATA"}}
      ]
    }],
    "generationConfig": {
      "responseModalities": ["TEXT", "IMAGE"]
    }
  }'

Configuration Options

Parameter Values Notes
aspectRatio 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9 Match use case
imageSize 1K, 2K, 4K Use uppercase K; PRO model only for 4K

Google Search Grounding (Real-time Data)

Add "tools": [{"google_search": {}}] to generate images based on current information:

curl -s -X POST \
  "https://generativelanguage.googleapis.com/v1beta/models/gemini-3-pro-image-preview:generateContent" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{"parts": [{"text": "Create an infographic of current tech stock prices"}]}],
    "tools": [{"google_search": {}}],
    "generationConfig": {
      "responseModalities": ["TEXT", "IMAGE"],
      "imageConfig": {"aspectRatio": "16:9"}
    }
  }'

Workflow

Step 1: Craft the Prompt

Use the ICS framework and prompting guide. Examples:

Photorealistic:

A photorealistic close-up portrait of an elderly Japanese ceramicist with deep wrinkles and a warm smile, inspecting a tea bowl. Soft golden hour light from a window. 85mm lens, shallow depth of field. Serene mood.

Infographic:

Create a clean, modern infographic explaining photosynthesis as a recipe. Show "ingredients" (sunlight, water, CO2) and "finished dish" (energy). Style like a colorful kids' cookbook page.

Product Shot:

High-resolution studio photograph of a matte black ceramic coffee mug on polished concrete. Three-point softbox lighting, 45-degree angle, sharp focus on rising steam. Square format.

Step 2: Generate Image

Use scripts/generate-image.sh or call API directly:

./scripts/generate-image.sh "Your prompt here" output.png --ratio 16:9 --size 2K

Step 3: Process Response

The API returns base64-encoded image data. Extract and decode:

# Response contains: {"candidates":[{"content":{"parts":[{"inlineData":{"mimeType":"image/png","data":"BASE64..."}}]}}]}
# Extract with jq and decode:
cat response.json | jq -r '.candidates[0].content.parts[] | select(.inlineData) | .inlineData.data' | base64 -d > image.png

Common Use Cases

Landing Pages & Ads

  • Use 16:9 or 21:9 for hero images
  • Specify brand colors, modern/minimal style
  • Include text requirements in prompt

Logos & Icons

  • Use 1:1 aspect ratio
  • Request "minimalist", "clean lines", "vector-style"
  • Specify color scheme explicitly

Product Photography

  • Describe lighting setup (softbox, natural, studio)
  • Mention surface/background materials
  • Include camera angle and lens type

Infographics

  • Define data to visualize
  • Specify style (corporate, playful, technical)
  • Request clear text and labeled sections

Stickers & Illustrations

  • Request "bold outlines", "kawaii", "cel-shading"
  • Specify "white background" or "transparent background"
  • Define color palette

Character Consistency (Multiple Images)

  • PRO supports up to 14 reference images
  • Explicitly state: "Keep facial features exactly the same as Image 1"
  • Describe expression/pose changes while maintaining identity

Scripts

See scripts/generate-image.sh for a ready-to-use generation script.

Detailed Prompting Guide

For advanced techniques including:

  • Photorealistic scene templates
  • Text rendering best practices
  • Sequential art and storyboarding
  • Dimensional translation (2D↔3D)
  • Search grounding for real-time data

Read references/prompting-guide.md.