Lumela: screenshot tool in Tauri 2, OCR goes through the OS engines instead of a bundled model #16508
tw4
started this conversation in
Show and tell
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
I made Lumela mostly for work. I share a lot of screenshots with other teams and they're full of emails, internal IPs and tokens, so it blurs those automatically. It also does the background/padding/shadow thing, and you can copy text out of a shot.
Some Tauri bits that might be useful to someone:
OCR uses whatever the OS has. Windows.Media.Ocr on Windows, Vision on macOS, tesseract on Linux. Since there's no model to ship, the Windows installer ended up around 4 MB and the dmg around 12.
The blurring is just Rust on top of the OCR output. Emails and IPs get parsed, card numbers have to pass Luhn, API keys are a regex plus an entropy check. Each hit returns the word boxes to cover.
Editor is React + Konva in the webview. Updates go through the Tauri updater with signatures. For Linux I made the .deb and .rpm depend on tesseract so OCR works without extra steps.
Repo: https://github.com/tw4/lumela (GPL-3.0)
Ask away if you want to know how something's wired up.
All reactions