Shotnest · Library
Review extracted text against the image
Before searching copied screenshot text, compare names, codes and amounts with the image; the Shotnest web tools do not read images themselves.
Where the text comes from
In the Shotnest app, text is read from each screenshot on the device by Apple’s text recognition, and the detail view offers Copy text with that result. On the website, nothing reads images: the text tools only work with what you paste. Apple Support’s page on Live Text, linked below, describes another way to copy text from a photo on iPhone and iPad.
Whatever the route, recognised text can drop a symbol, merge two columns into one line or confuse similar shapes such as 0 and O or 1, l and I. Search and comparison then work on the copy, not on the image.
Check a copy before using it
- Copy only the part you need for the search or comparison.
- Compare every name, number and code in the copy with the image.
- Keep ambiguous characters as they are and mark them for review instead of guessing.
- Check line breaks where the screenshot had columns or a table.
- Paste the checked text into the tool.
One misread character
A record copied as “Ref 0O-1lI” and searched as “ref 00-11i” returns no match. Normalization lowercases and folds accents, but it does not turn the letter O into the digit 0 or l into 1. The difference has to be fixed by looking at the image.
- Record
- Ref 0O-1lI
- Query
- ref 00-11i
- Matching records
- none
A lost currency symbol changes classification too. “Total €14.25” and “Paid €14.25” score 3 for Receipts, which is confident; the same lines copied without the euro sign score 2, because the rules no longer count two prices.
- With € · Receipts
- score 3
- Without € · Receipts
- score 2
Extraction errors to expect
- A currency symbol lost from a price, which also removes it from the price count.
- Two table columns joined so that amounts move next to the wrong labels.
- Hyphenated words split across lines.
Review stays with you
The website’s text tools do not perform OCR, and the result of any recognition still needs a check against the image before you rely on it.