Sniper region selection
Select a subtitle, dialog, or UI area once and let USTA keep monitoring it.
Real-time OCR translation for Linux desktops
USTA watches the exact area you choose, extracts text with OCR, translates it instantly, and places the result back on your screen as a focused HUD overlay.
USTA combines region capture, OCR, machine translation, and overlay rendering into a lightweight desktop companion for real-time language support.
Select a subtitle, dialog, or UI area once and let USTA keep monitoring it.
Use OCR engines such as Tesseract, EasyOCR, RapidOCR, or PaddleOCR depending on your setup.
Translate recognized text through Google Translate or DeepL-compatible engines.
Keep translated text visible above the original content without switching windows.
Broadcast every translation as JSON for OBS, logging, subtitles, TTS, or custom tools.
Optimized for modern Linux desktops, with current focus on KDE Plasma and GNOME Wayland workflows.
Designed for moments when pausing, copying, or switching applications breaks the flow: visual novels, RPGs, videos, livestreams, tutorials, and foreign-language software. Open the demo on YouTube.
Follow dialog and UI text in untranslated titles.
Translate subtitles or captions as they appear.
Pipe translation output into overlays or OBS scenes.
Send recognized text to logs or text-to-speech tools.
USTA keeps the workflow simple, repeatable, and script-friendly.
Use sniper mode to capture only the part of the screen that matters.
The selected engine extracts text continuously from the region.
The translated result appears in the overlay and is also broadcast over IPC.
Every translation can be streamed from the local Unix socket at /tmp/usta.sock.
Current builds are optimized for KDE Plasma and GNOME Wayland environments using xcb-compatible workflows. Wider platform support is planned.
Explore the source, follow setup instructions, and try USTA on your Linux desktop.