Re-read a selected region with PaddleOCR-VL, served by the Local OCR app
You can now re-read just the region you select on the image (a single line, say) with PaddleOCR-VL. It runs on your own computer and the image never leaves it.
How: (1) Install and open Local OCR (https://lo.ldas.jp) and turn on "Let other tools use this" in its settings. (2) In this editor, choose "PaddleOCR-VL (runs on this computer)" as the OCR engine. (3) Select a region, or draw one with "Add region", and press "Run OCR on this region". The line text is replaced with the new reading.
PaddleOCR-VL reads one line at a time, so it cannot do full-page OCR. Create the regions with NDL Koten OCR first.
The older launcher (zip) still works.
Folders with more than 1,000 files were listed only up to the 1,000th
File listings stopped at the 1,000th entry in a folder. Anything after that never appeared, no matter how far you scrolled.
GitHub returns at most 1,000 entries per folder and drops the rest silently, so the truncated list looked complete.
Folders holding more than 1,000 files now use a different call that returns every entry, so the previously hidden files can be opened from the list.
Searching by part of a file name has always worked and was never affected. On large folders it is still the quickest way to reach a file.
New regions now insert their line right after the selected line
Previously, drawing a new region always appended its transcription line at the end of the list, which meant scrolling back on long pages.
New lines are now inserted right after the line you last worked at (the clicked line, the line of the selected rectangle, or the line you just created). Consecutive regions keep their drawing order. With nothing selected, lines are appended at the end as before.
Region number labels no longer overlap the text on the image
The "#N" label is now placed by the rectangle's orientation: above tall rectangles (vertical columns) and to the left of wide ones (horizontal lines). Also fixed deleted regions occasionally leaving their rectangle on the image.
Every region now appears as a row in the transcription list
The separate zones column is gone: every region on the image is a row in the transcription list. Untranscribed regions show as empty lines you can fill in place by typing or OCR, and are highlighted in red on the image.
In-browser OCR (per region and full page)
OCR now runs entirely in your browser — no image leaves your machine. Recognize a single region or detect-and-recognize a whole page, with switchable engines.
