Generate Image

Nano Banana Image Generation

Generate professional images via the Gemini API's native image generation models.

When to Use This Skill

ALWAYS use this skill when the user:

• Asks for any image, graphic, illustration, or visual
• Wants a thumbnail, featured image, or banner
• Requests icons, diagrams, or patterns
• Asks to edit, modify, or restore a photo
• Uses words like: generate, create, make, draw, design, visualize

Do NOT attempt to generate images through any other method.

Prerequisites

Verify API key is set:

[ -n "$GEMINI_API_KEY" ] && echo "API key configured" || echo "Missing GEMINI_API_KEY"

If missing, ask the user to set it:

export GEMINI_API_KEY="your-key-from-aistudio.google.com/apikey"

Models

Default model: gemini-2.5-flash-image for most requests. Use gemini-3-pro-image-preview when user needs: 4K resolution, accurate text in images, complex multi-element compositions, professional asset production, or Google Search grounding.

Image Generation (Text-to-Image)

curl -s "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-image:generateContent?key=$GEMINI_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
    "contents": [{"parts": [{"text": "YOUR PROMPT HERE"}]}],
    "generationConfig": {
      "responseModalities": ["TEXT", "IMAGE"]
    }
  }' | python3 -c "
import json, sys, base64
data = json.load(sys.stdin)
if 'error' in data:
    print('API Error:', data['error'].get('message', str(data['error'])))
    sys.exit(1)
candidates = data.get('candidates', [])
if not candidates:
    print('No candidates returned')
    sys.exit(1)
parts = candidates[0].get('content', {}).get('parts', [])
for part in parts:
    if 'inlineData' in part:
        img_data = base64.b64decode(part['inlineData']['data'])
        mime = part['inlineData'].get('mimeType', 'image/png')
        ext = 'png' if 'png' in mime else 'jpg'
        outpath = 'OUTPUT_PATH_HERE.' + ext
        with open(outpath, 'wb') as f:
            f.write(img_data)
        print(f'Image saved to: {outpath} ({len(img_data)} bytes)')
    elif 'text' in part and not part.get('thought'):
        print('Model notes:', part['text'][:300])
"

With Aspect Ratio

Add imageConfig to control aspect ratio:

curl -s "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-image:generateContent?key=$GEMINI_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
    "contents": [{"parts": [{"text": "YOUR PROMPT HERE"}]}],
    "generationConfig": {
      "responseModalities": ["TEXT", "IMAGE"],
      "imageConfig": {
        "aspectRatio": "16:9"
      }
    }
  }' | python3 -c "SAME_DECODER_SCRIPT"

Image Editing (Image + Text to Image)

To edit an existing image, base64-encode it and include it as an inlineData part:

python3 -c "
import base64, json, subprocess, sys

with open('INPUT_IMAGE_PATH', 'rb') as f:
    b64 = base64.b64encode(f.read()).decode()

payload = {
    'contents': [{'parts': [
        {'inlineData': {'mimeType': 'image/png', 'data': b64}},
        {'text': 'YOUR EDIT INSTRUCTION HERE'}
    ]}],
    'generationConfig': {
        'responseModalities': ['TEXT', 'IMAGE']
    }
}

result = subprocess.run(
    ['curl', '-s',
     'https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-image:generateContent?key=' + '$GEMINI_API_KEY_VALUE',
     '-H', 'Content-Type: application/json',
     '-d', json.dumps(payload)],
    capture_output=True, text=True
)

data = json.loads(result.stdout)
if 'error' in data:
    print('API Error:', data['error'].get('message', str(data['error'])))
    sys.exit(1)
for part in data.get('candidates', [{}])[0].get('content', {}).get('parts', []):
    if 'inlineData' in part:
        img_data = base64.b64decode(part['inlineData']['data'])
        with open('OUTPUT_PATH', 'wb') as f:
            f.write(img_data)
        print(f'Image saved ({len(img_data)} bytes)')
    elif 'text' in part and not part.get('thought'):
        print('Model notes:', part['text'][:300])
"

Aspect Ratios

Prompt Tips

1. Describe the scene narratively — a descriptive paragraph beats a tag list
2. Add "no text" if you don't want text rendered in the image
3. Be hyper-specific — instead of "fantasy armor" say "ornate elven plate armor, etched with silver leaf patterns"
4. Use photography terms for photorealistic results — camera angles, lens types, lighting
5. Reference artistic styles — "editorial photography", "flat illustration", "3D render", "watercolor"

Presenting Results

After generation:

1. Read the generated image file to show it to the user
2. Offer to regenerate with variations or different models if needed

Source: ash-ai-org/ash-ai — distributed by TomeVault.