EN | RU

Advanced screen translator. Translumo is able to detect and translate appearing in the selected area text in real-time (e.g. subtitles).

Main features

High text recognition precision

Translumo allows to combine the usage of several OCR engines simultaneously. It uses machine learning training model for scoring each recognized result by OCR and chooses the best one.

Simple interface
Low latency
Integrated modern OCR engines: Tesseract 5.2, WindowsOCR, EasyOCR
Available translators: Google Translate, Yandex translate, Naver Papago, DeepL
Available recognition languages: English, Russian, Japanese, Chinese (simplified), Korean
Available translation languages: English, Russian, Japanese, Chinese (simplified), Korean, French, Spanish, German, Portuguese, Italian

System requirements

Windows 10 build 19041 (20H1) / Windows 11
DirectX11
8 GB RAM (for mode with EasyOCR)
5 GB free storage space (for mode with EasyOCR)
Nvidia GPU with CUDA SDK 11.8 support (GTX 7xx series or later) (for mode with EasyOCR)

How to use

Open the Settings
Select Languages->Source language and Languages->Translation language
Select Text recognition->Engines (please check Usage tips for recommendation modes)
Select capture area
Run translation

Usage tips

Generally, I recommend always keep Windows OCR turned on. This is the most effective OCR for the primary text detection with less impact on performance.

Recommended combinations of OCR engines

Tesseract-Windows OCR-EasyOCR - advanced mode with the highest percision
Tesseract-Windows OCR - noticeably less impact on system performance. It will be enough for cases when text has simple solid background and font is quite common
Windows OCR-EasyOCR - for very specific complex cases it makes sense to disable Tesseract and avoid unnecessary text noises

Select minimum capture area

It reduces chances of getting into the area random letters from background. Also the larger frame will take longer to process.

Use proxy list to avoid blocking by translation services

Some translators sometimes block client for a large number of requests. You can configure personal/shared IPv4 proxies (1-2 should be enough) on Languages->Proxy tab. The application will alternately use proxies for requests to reduce number from one IP address.

Use Borderless/Windowed modes in games (not Fullscreen)

It is necessary to display the translation window overlay correctly.

If the game doesn't have such mode, you can use external tools to make it borderless (e.g. Borderless Gaming)

Install the application on SSD

To reduce cold launch time with enabled EasyOCR engine (loading large EasyOCR model into RAM).

FAQ

I got error "Failed to capture screen" or nothing happens after translation starts

Make sure that target window with text is active. Also try to restart Translumo or reopen target window.

I got error "Text translation is failed" after successful using the translation

There is a high probability that translation service temporarily blocked requests from your IP. You can change translator or configure proxy list.

Can't enable Windows OCR

Make sure that the application is runned as an Administrator. Translumo each time tries check installed Windows language pack via PowerShell.

I set borderless/windowed mode, but a translation window is still displayed under a game

When game is running and focused use the hotkey (ALT+T by default) to hide and then show again translation window

Package downloading for EasyOCR failed

Try to re-install it under VPN

Hotkeys don't work

Other applications may intercept specific hotkeys handling

I got error "Text detection is failed (TesseractOCREngine)"

Make sure that the application path contains only latin letters.

Build

Visual Studio 2022 and .NET 7 SDK are required

Clone repository (master branch always equals last release version):

git clone https://github.com/Danily07/Translumo.git

Restore packages and build a solution. binaries_extract.bat will be executed during building, which will automatically download models and python binaries to target output directory.

olegafs / translumo Goto Github PK

translumo's Introduction