xulihang Well-Known Member Licensed User Longtime User Jun 22, 2026 #1 I've developed a browser extension for image translation with the help of AI: https://github.com/xulihang/imagetrans_chrome_extension. Now I want to create an Android version to translate the screen. Can I use B4A to develop such an app? It needs to do the following: 1. Take a screenshot of the current screen. 2. Show the translation above another app via overlay or a dialog. 3. On-device OCR, translation, text removing and rendering with OpenCV and ONNXRuntime. (Optional to call a B4J server app for better performance) 4. Provide a translation API for the browser extension to use.
I've developed a browser extension for image translation with the help of AI: https://github.com/xulihang/imagetrans_chrome_extension. Now I want to create an Android version to translate the screen. Can I use B4A to develop such an app? It needs to do the following: 1. Take a screenshot of the current screen. 2. Show the translation above another app via overlay or a dialog. 3. On-device OCR, translation, text removing and rendering with OpenCV and ONNXRuntime. (Optional to call a B4J server app for better performance) 4. Provide a translation API for the browser extension to use.
mcqueccu Expert Licensed User Longtime User Jun 22, 2026 #2 Yes, I think it's should be fairly moderate to achieve Use snapshot, Bxdialog, OCR Upvote 0
xulihang Well-Known Member Licensed User Longtime User Monday at 5:37 AM #3 I ended up using Java + XML with the help of AI agents to create such a project within two weeks: https://github.com/xulihang/ImageTrans-Android Using B4A would be troublesome for such an application. Upvote 0
I ended up using Java + XML with the help of AI agents to create such a project within two weeks: https://github.com/xulihang/ImageTrans-Android Using B4A would be troublesome for such an application.
hatzisn Expert Licensed User Longtime User Monday at 6:43 AM #4 I think it would be much easier to use Google Lens for a B4X Mobile Project (in conjunction with Google translate API). Also Google Translate App for mobile phones already does this... The solution with Google Lens would be as easy as: 1) Notify the user that you will open Google Lens App and that he should take the photo where the text appears and copy the text. 2) The user must get back to your app 3) Paste the text into your app or take it directly from the clipboard and use Google Translate API to translate it. Also last time I checked Google Lens, it has also a translate button although I did not try it as I did not have any foreign text available around me. Last edited: Monday at 6:58 AM Upvote 0
I think it would be much easier to use Google Lens for a B4X Mobile Project (in conjunction with Google translate API). Also Google Translate App for mobile phones already does this... The solution with Google Lens would be as easy as: 1) Notify the user that you will open Google Lens App and that he should take the photo where the text appears and copy the text. 2) The user must get back to your app 3) Paste the text into your app or take it directly from the clipboard and use Google Translate API to translate it. Also last time I checked Google Lens, it has also a translate button although I did not try it as I did not have any foreign text available around me.
xulihang Well-Known Member Licensed User Longtime User Monday at 7:18 AM #5 On-device model running with ONNXRuntime is better than Google Lens for my case, like manga translation. Upvote 0
On-device model running with ONNXRuntime is better than Google Lens for my case, like manga translation.
epiCode Active Member Licensed User Monday at 8:54 AM #6 @xulihang Are you trying to do something on the lines of MS PC Manager ? Upvote 0
hatzisn Expert Licensed User Longtime User Monday at 9:15 AM #7 xulihang said: On-device model running with ONNXRuntime is better than Google Lens for my case, like manga translation. Click to expand... Codex has proposed the ONNX runtime for a wake word, when I was searching for this matter. Upvote 0
xulihang said: On-device model running with ONNXRuntime is better than Google Lens for my case, like manga translation. Click to expand... Codex has proposed the ONNX runtime for a wake word, when I was searching for this matter.