Check a PDF transcript extraction before pasting it into your editor

A PDF-to-text command can finish successfully and still leave you with an empty file or dialogue in the wrong order. Before pasting a PDF interview draft into an editor, check those two outcomes separately. Our image-only control produced no words, while our selectable two-column draft produced all the words in an unexpected sequence.

Make two extraction copies

Keep the original PDF unchanged and choose new output names. With Poppler’s pdftotext installed, run:

pdftotext interview.pdf default.txt
pdftotext -layout interview.pdf layout.txt

Open both text files alongside the PDF. Default output is a useful prose starting point; layout mode can make columns and spacing easier to compare. Neither is a guarantee that the intended reading order has been reconstructed.

Reject an output with no meaningful text

Our scan control returned success but contained just a form-feed character. A file-size check alone would accept that byte. This Python 3 check rejects whitespace-only output:

from pathlib import Path

text = Path("default.txt").read_text(encoding="utf-8")
if not text.strip():
    raise SystemExit("No meaningful extracted text; inspect the PDF text layer.")
print("Text exists; sequence still needs review.")

This is a minimal gate, not an OCR detector. An unrelated heading might make the file nonempty while the interview itself is still an image. Check a sentence from the start, middle and end of the draft. If those words are absent, request a text export or use an authorized OCR workflow; do not fill the gaps from memory.

Check sequence with the actual source

Our four timestamped test lines came out as 00:01, 00:03, 00:02, 00:04 in default mode. Layout mode kept left and right text paired horizontally. That made the two-column structure visible but did not convert it into chronological dialogue.

Compare speaker labels and complete sentences against the PDF. Do not sort every transcript line by a timestamp automatically: timestamps may repeat, wrap onto another line or refer to different sections. Resolve the document’s intended reading order before editing or constructing subtitles.

Remove page markers only after checking boundaries

If your destination needs a continuous plain-text file, make another extraction with explicit suppression of page breaks:

pdftotext -nopgbrk interview.pdf continuous.txt

Our two-page fixture then contained no form-feed characters. This option did not change the problematic dialogue ordering. Check where the pages meet so that a header, footer or sentence fragment is not mistaken for transcript content.

Know when to stop

This workflow does not create SRT cue timing, identify speakers automatically, perform OCR or verify speech against audio. It gives you inspected editable text. Keep a source copy and review against the recording before publication. For the tested results and tool tradeoffs, read our pdftotext review and the command manual.

Tests used pdftotext 24.02.0 on Ubuntu with a synthetic two-page text PDF and a separate image-only PDF. The empty-output check was also run against these results. No protected, large, mobile or browser cases were tested.