Getting the text out of five screenshots inside one PDF
The document is a three-page weekly ops report with five screenshots pasted in: a growth dashboard, a chat channel, a spreadsheet, an alert and a kanban board. I built it myself as a test file with invented content, because it's such a common shape of document: someone screenshots five tools and drops them into a report, and later someone else needs the words. Select-all and copy doesn't get you…
A three-page weekly operations report contained five screenshots - a growth dashboard, chat channel, spreadsheet, alert, and kanban board. The document was manually created for demonstration purposes, as combining screenshots of five tools into a report is a common format. Selecting and copying text from the screenshots did not yield the desired result, as the PDF treats the screenshots as images without text.
To extract the text, the entire PDF was uploaded to ImgIn's Extract PDF content tool. The OCR (Optical Character Recognition) feature recognized all five screenshots in a single run, placing each as a separate block labeled "Image OCR" immediately after its corresponding screenshot and before its caption. The order of the blocks matched the page order of the original document.
The extracted text for each screenshot was accurate and well-formatted. The growth dashboard's data appeared correctly in tabular form, with labels, values, and deltas preserved. The chat messages were correctly attributed to the right speakers, except for two instances where the names of the speakers were incorrectly associated with the wrong messages.
The spreadsheet data was perfectly extracted and tab-separated, making it easy to paste directly into a spreadsheet application. The alert's text was also accurately extracted, with the exception of a formatting issue where the title lost spaces around its dot.
Two screenshots had less satisfactory results. A bar chart under the dashboard tiles had seven values that were correctly extracted but incorrectly grouped into two rows due to the OCR process. Similarly, a kanban board had three column names and three of the four cards correctly extracted. However, the fourth card, OPS-416, was incorrectly extracted as just its ID, with its title missing from the output.
Additionally, the small headers and footers that appeared on each page of the original report were not included in the extracted text.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
