
whisk
by AndyShaman
Generate images with Google Whisk AI (Imagen 3.5) directly from Claude Code
SKILL.md
name: whisk description: Use when user asks to generate, create, or draw an image, photo, picture, logo, illustration, or art. Triggers include 'сгенерируй', 'нарисуй', 'создай картинку', 'сделай фото', 'фотографию', 'изображение', 'иллюстрацию', 'логотип', 'generate image', 'create image', 'draw', 'make a picture', 'visualize'.
Whisk - Image Generation via Google Imagen 3.5
Generate images using Google Whisk AI directly from Claude Code.
When to Use
- User asks to generate/create/draw an image, photo, picture
- User wants a logo, illustration, art, or visualization
- Keywords: "нарисуй", "сгенерируй", "создай картинку", "сделай фото", "generate", "draw", "create image"
Workflow (IMPORTANT)
Before generating an image, ALWAYS ask the user:
- Output folder — Where to save the image?
- Default:
./whisk-images - User may specify a different folder
- Default:
Use AskUserQuestion tool with options:
- "По умолчанию (whisk-images)" / "Default (whisk-images)"
- "Указать другую папку" / "Specify another folder"
If user chooses default or doesn't respond, use ./whisk-images.
Context Optimization (CRITICAL)
NEVER read generated images with the Read tool!
Why: Reading PNG files converts them to base64 and consumes thousands of tokens from context window.
After generation:
- Show only the file path to user
- Tell user: "Изображение сохранено: "
- DO NOT use Read tool on generated .png files
- DO NOT display the image inline
User can open the file manually in their file manager or image viewer.
Quick Reference
| Command | Description |
|---|---|
whisk generate "prompt" | Generate image |
whisk generate "prompt" -c 3 | Generate 3 variants |
whisk generate "prompt" -r 16:9 | Landscape format |
whisk generate "prompt" -o ./folder | Save to folder |
whisk status | Check auth status |
| Ratio | Size | Use Case |
|---|---|---|
1:1 | 1024x1024 | Avatars, icons, logos |
16:9 | 1365x768 | Banners, landscapes |
9:16 | 768x1365 | Stories, posters |
CLI Usage
whisk generate "your prompt" [options]
Options
-c, --count <n>— Number of images (1-10), default: 1-r, --ratio <ratio>— Aspect ratio (1:1, 16:9, 9:16), default: 1:1-o, --output <dir>— Output directory, default: ./whisk-images
Examples
# Simple generation
whisk generate "cat in space, digital art"
# Multiple variants
whisk generate "minimalist coffee shop logo" -c 3
# Landscape banner
whisk generate "sunset over ocean" -r 16:9
# Custom output folder
whisk generate "abstract background" -o ./assets/bg
Authorization Flow
CLI handles authorization automatically:
- If no token exists, CLI starts auth-server on port 3847
- CLI shows instructions to user
- User opens labs.google/fx/tools/whisk in Chrome
- User clicks Connect in Whisk Proxy extension
- Extension sends token to localhost:3847
- CLI receives token and continues generation
No manual server startup required!
Prompt Tips
- Be specific: "orange cat on windowsill" > "cat"
- Add style: "digital art", "watercolor", "photorealistic", "minimalist"
- Composition: "centered", "close-up", "wide angle"
- Atmosphere: "cozy", "dramatic lighting", "vibrant colors"
Troubleshooting
"Not authorized" / "Не авторизован"
CLI will automatically show authorization instructions. Follow them:
- Open https://labs.google/fx/tools/whisk in Chrome
- Log into Google account
- Click Connect in Whisk Proxy extension
"Token expired"
CLI will automatically re-request authorization when token expires.
Check status
whisk status
Score
Total Score
Based on repository quality metrics
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
1ヶ月以内に更新
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
Reviews
Reviews coming soon
