Image Text Recognition
Installs in TopArrow. If you do not have it yet, setup opens this app.
Install “Image Text Recognition” in TopArrow
Opening TopArrow. If it does not open, download TopArrow. After setup it opens this app’s install page.
Download TopArrowAlready have TopArrow? Search for this code in the Market.
top3262668813-5mbtkwWhat this version does
Image Text Recognition
Turn text in images into editable, copyable text. The app includes an offline OCR engine and Simplified Chinese and English language models, so you can start without an account or an external service.
Features
- Flexible image import: Select a file, drag and drop an image, or explicitly paste an image from your clipboard.
- Offline OCR: Recognize Simplified Chinese, English, or mixed Chinese and English using the included engine and language models.
- Edit and export: Edit recognized text, copy it, or save it as a TXT file.
- Connect your own AI: Use an OpenAI-compatible vision Chat Completions endpoint with your API URL, model, key, and recognition prompt.
- Custom OCR APIs: Send images as Base64 JSON, Data URL JSON, or multipart file uploads. Configure the request template, file field, additional headers, and response text field.
- Multiple API profiles: Add, edit, delete, and test configurations with the current image. Requests support timeouts and cancellation.
Privacy and network use
Local OCR works offline and does not automatically upload images. Clipboard images are read only when you explicitly request pasting; the app does not continuously monitor the clipboard.
Before an AI request, custom API request, or connection test sends an image, the app displays the destination and asks for confirmation. After confirmation, the current image is sent to your configured service. Public endpoints use HTTPS. HTTP connections to a service you run locally require an explicit opt-in and are limited to loopback addresses. The app does not expose a LAN or public server and does not use the microphone or camera.
The app provides connection tools only. It does not supply cloud OCR, AI accounts, hosted models, or API credits. Third-party services may charge fees; their capabilities, pricing, and data handling depend on the provider you choose. API keys are stored using system encryption when available; otherwise, they remain in memory only until the app exits.How to use
- Select, drop, or explicitly paste an image.
- Choose local OCR and a recognition language. To use an external service, save your own API configuration in Settings and select it as the recognition method.
- Start recognition, then edit, copy, or export the text.
Current limitations
- Supported formats: PNG, JPEG, WebP, and BMP. Each image can be up to 20 MB and 40 megapixels.
- The built-in screen capture entry is currently unavailable. Take a screenshot with your system tools, then paste it into the app.
- Results depend on image clarity, text size, font, and layout. Review handwriting, tilted text, and complex layouts manually. No fixed accuracy rate is promised.
- External services must support image input and a compatible request format. Custom APIs currently use POST with a single JSON response and a string text field. Streaming, asynchronous polling, and dynamic request signing are not supported.
- The user interface is in Chinese.