gemini-image-translator/SKILL.md

---
name: gemini-image-translator
description: Translate text inside images with Gemini 2.5 Flash, then render the translated text back onto the original image. Use this skill when given a local image path and an optional target language.
---

# gemini-image-translator

Translate text in a local image with Gemini 2.5 Flash OCR + translation, then write the translated text back onto the image with `sharp`.

Set `GEMINI_API_KEY` in the environment before running.

## Run

```bash
bun scripts/run.ts translate <image-path> [target-language] [output-path] [--dry-run]
```

## Commands

| Command | Description |
|---------|-------------|
| `translate <image-path> [target-language] [output-path]` | OCR the image, translate all detected text to the target language (default `zh-CN`), and save a new image with translated text overlaid. |

## Output

Returns JSON with `status`, `command`, `dryRun`, and `data`.

On success, `data` includes:

- `inputPath`: resolved source image path
- `outputPath`: generated output image path
- `targetLanguage`: requested target language
- `width` / `height`: image dimensions
- `regions`: detected OCR regions with translated text and pixel bounds