This is a browser OCR app built using NextJS that allows you to use Google Cloud Vision to automatically convert images from your screen into editable/scannable text
You will need Docker and a Google Service Account key with Cloud Vision API access. The API key can be created in the Google Cloud Console and every user gets 1000 free Cloud Vision API requests per month, and after that it's $1.50 per 1000 as of August 2024.
- Clone this repository
- Copy your Google JSON key into
./config/google_service_account.json - Run
docker compose -f docker-compose.dev.yml up - Open
http://localhost:3001and click the "Select screen" button (you may have to scroll down) - Select a screen and you will be able to draw a rectangle around whatever you want to run OCR on
The tool will automatically update if it detects a change in content on the screen, populating the results box below your stream. If you find that it is refreshing too often due to dynamic backgrounds, you can use the filters at the bottom of the page. There is a preview for the filter result underneath the filter settings. Try to adjust it so that only the text is visible; this will minimize the number of false-positive detections and minimize your OCR API usage.
If the tool does not detect a change in the selected image but you want to force it to process the data, click the "Force process image" button and it will override the automatic process to detect changes and force an immediate OCR request.
If you want to pause the capture loop, click the "Stop capture loop" button. This is useful if you're going to a menu or somewhere that you don't want to waste OCR requests on. You can resume the image capture with the "Start capture loop" button.