OCR Overview
Tex extracts text from any captured image using Tesseract 5.5.1, wrapped behind the IOcrService interface. Any image loaded in the editor can be turned into editable text, along with a confidence score and processing time.
Two Entry Points
| Entry point | Hotkey | Behaviour |
|---|---|---|
| Region with OCR | Ctrl+Shift+O | Opens the selection overlay, captures the region, and runs OCR in one step. No editor stage — result popup appears directly. |
| Extract Text | Ctrl+T | Runs OCR on the capture currently loaded in the main editor. Button is disabled when no capture is loaded. |
The first method is best for quick text grabs from screen content. The second is for captures you already have open (possibly after annotating or cropping).
Result Shape
Every OCR call returns an OcrResult with three values:
| Field | Unit | Typical range |
|---|---|---|
| Text | plain string | any length |
| Confidence | percent | 0–100 |
| Processing time | milliseconds | 50 ms (small region) to 3 s (full screen, CJK language) |
Both entry points display these in the OCR Result Popup (see 04.03).
Lazy Initialization
Tex does not load the Tesseract engine at startup — it loads on first use. Expect a one-time delay of 2–5 seconds on the first OCR of each session as the language data loads into memory. Subsequent calls are near-instant.
Tip — If you know you will use OCR, trigger one extraction early in your session to pay the warm-up cost before you need fast results.
What's Next
04.02— supported languages and the preprocessing toggle.04.03— anatomy of the result popup.04.04— OCR history window.