QuickOCR 1.1.0

This release fixes the problems reported against 1.0.1, including the one where the screen
appeared to freeze. If you are on 1.0.1, please update.

The frozen screen

The capture overlay covered the whole desktop, stayed on top of everything, and had no working
way out — Escape did nothing, because the overlay never actually received keyboard focus.
Anything you opened to escape the situation, Task Manager included, opened behind it. If the
app stalled while the overlay was up, Windows replaced it with a blank white copy of the window.

Now:

• Escape works. So does right-click, and a single click without dragging.
• The overlay gets out of the way the moment another window takes over, so nothing is ever
  trapped behind it.
• It closes itself after two minutes no matter what.
• Every way a capture can end now closes the overlay and gives you the app back.

Text that came out wrong, or never arrived

• Grey and low-contrast text was being destroyed before Tesseract ever saw it. Anything
  lighter than about #5a5a5a was silently garbled or erased entirely — Balance 1,024.50 USD
  came back as 2Paance 10e4.50 JSS, or as nothing at all. QuickOCR now tries several
  approaches and keeps whichever one Tesseract is most confident about.
• "Copied to clipboard" was claimed whether or not it worked. The text is now read back and
  checked before anything is claimed, and you get a clear error if it failed.
• Low-confidence results are flagged instead of being presented as a clean read, so you know
  to check before pasting.
• The overlay's own dimming could end up baked into the capture, throwing the result off. It is
  now hidden before the screen is grabbed.
• Multi-monitor setups with different scaling captured the wrong part of the screen. Fixed.
• Reading no longer freezes the window — it happens in the background.

Languages

Text in other languages came out mangled — German Größe read as GroBe, Spanish niño as
nino. That was never a text-encoding problem: QuickOCR simply only had English and French
recognition data.

• There is now a Language picker in the window, and your choice is remembered.
• You can add any language yourself: drop the .traineddata file from
  tessdata_fast into Tesseract-OCR/tessdata/.
  Nothing else to change. See the README.
• This build still ships English and French only.

Tesseract is most accurate with a single language selected, so prefer one over a combination.

Smaller download

46 MB, down from 84 MB. The old build carried the entire Tesseract distribution — 22 unused
libraries, 17 training tools, an uninstaller and some Java components that the app never touched.
Startup is quicker too, since all of that was being unpacked every time you launched it.

Reporting problems

QuickOCR now keeps a log at %APPDATA%\QuickOCR\quickocr.log. Attaching it to an issue makes a
problem far easier to track down — the previous version recorded nothing at all.

---

Windows 10/11, no installation, no admin rights. Tesseract is bundled.