localize-screenshots
Translate text in app screenshots to multiple languages using AI, supporting phone and tablet images for international app store optimization.
Instructions
Translate app screenshots to multiple languages using Gemini API.
IMPORTANT: This tool uses the search-app tool internally to validate the app. You can provide an approximate name, bundleId, or packageName.
This tool:
Validates the app exists in registered-apps.json
Reads supported locales from public/products/{slug}/locales/ directory
Scans screenshots from the primary locale's screenshots folder
Uses Gemini API (imagen-3.0-generate-002) to translate text in images
Validates output image dimensions match source and resizes if needed
Requirements:
GEMINI_API_KEY or GOOGLE_API_KEY environment variable must be set
Screenshots must be in: public/products/{slug}/screenshots/{locale}/phone/ and /tablet/
Locale files must exist in: public/products/{slug}/locales/
Example structure:
public/products/my-app/
├── config.json
├── locales/
│ ├── en-US.json (primary)
│ ├── ko-KR.json
│ └── ja-JP.json
└── screenshots/
└── en-US/
├── phone/
│ ├── 1.png
│ └── 2.png
└── tablet/
└── 1.pngInput Schema
| Name | Required | Description | Default |
|---|---|---|---|
| appName | Yes | App name, slug, bundleId, or packageName to search for. Will be validated using search-app. | |
| targetLocales | No | Specific target locales to translate to. If not provided, all supported locales from the product will be used. | |
| deviceTypes | No | Device types to process (default: both phone and tablet) | |
| dryRun | No | Preview mode - shows what would be translated without actually translating | |
| skipExisting | No | Skip translation if target file already exists (default: true) | |
| screenshotNumbers | No | Specific screenshot numbers to process. Can be: - Array for all devices: [1, 3, 5] - Object for per-device: { phone: [1, 2], tablet: [1, 3, 5] } If not provided, all screenshots will be processed. | |
| preserveWords | No | Words to keep untranslated (e.g., brand names, product names). Example: ["Pabal", "Pro", "AI"] |