Text recognition problems
Recognition uses Apple’s Vision framework, on your Mac. It is strong on screen text — high contrast, aligned, no skew — and weaker on photographs of screens and on handwriting.
“No text found” on something that clearly has text
Check the profile. Settings → Text Recognition → Profile. Automatic chooses per image using recognition confidence, but you can force Screen text or Handwriting & whiteboards.
If the capture is a photo of a whiteboard or a handwritten note, the handwriting profile will do noticeably better.
Check the language. Settings → Text Recognition → Languages lists the Vision languages your Mac has installed. If the screenshot is in a language that isn’t selected, recognition will do poorly or return nothing.
To add a language, install it in System Settings → General → Language & Region, then reopen Screenshot Inbox — the list is read from what macOS has available.
It may genuinely have no text. A chart with no labels, an icon, a photograph. The app marks this as a real state rather than an error, and the capture is still stored and filterable.
Recognition is queued but never finishes
Low Power Mode. Recognition pauses automatically to avoid being blamed for battery drain. Plug in, or turn off Low Power Mode, and it resumes.
A large import is ahead of it. Recognition works newest-first, so recent captures are done promptly, but a five-thousand-file backlog import takes a while to work through.
The text is recognized but wrong
Small, anti-aliased or low-contrast text is where Vision struggles — a terminal at 9pt on a dark theme, scaled down, is a hard case.
This is exactly why search is typo-tolerant: when exact matching finds nothing, a bounded edit-distance pass runs, so a character-level misread usually doesn’t hide the capture.
If a capture matters and was read badly, retaking it at a larger size will help more than anything in Settings.
Reprocessing
Settings → Text Recognition → Reprocess All re-runs recognition across your whole library. Use it after changing language or profile settings.
It takes about as long as the original import did and runs in the background. Existing tags and manual edits are preserved.
Line breaks and formatting
Recognized text preserves line breaks, which is what makes copying a stack trace or a block of code paste usefully.
Column layouts are read in the order Vision reports them, which for a two-column page is not always the reading order a human would choose.
Didn't answer it? Write tosupport@screenshotinbox.com — one person reads it, and questions that land here twice become a new page.