How to Tell If an Audio Interview Has Been Cut or Edited

Editing is normal. Knowing what got cut is the actual question

Almost every published interview is edited — for length, for clarity, to remove dead air. That’s not the issue. The issue is when you need to know specifically what was removed or reordered, because the edit changed meaning, or you’re fact-checking a quote against its original context, or you’re the interviewee and want to confirm the published version matches what you actually said and in what order. Listening back to both recordings side by side and trying to track where they diverge by ear is slow and easy to get wrong on anything longer than a couple of minutes.

Compare the raw and published versions

Upload your original recording and the published version to DiffALL’s audio comparison. You’ll get a similarity score, a per-second chart, and mel spectrograms — a visual representation of the audio’s frequency content over time, which makes structural changes like cuts and reorders visible rather than something you have to catch by ear.

A clean vertical break in the spectrogram at a specific timestamp, where the frequency pattern on one side has no relationship to the pattern on the other side at that point, is a cut. Compare that timestamp in the published version against your raw recording’s timeline — if you can identify what was playing in the raw version right before and after that point, you’ve found the edit.

A section of the published audio that matches a different timestamp in the raw recording than where it appears in your timeline — not missing, just moved — is a reorder. This is the case a straightforward listen is worst at catching, since both versions can sound completely natural in isolation while presenting statements in a different sequence than they were actually said.

A gap in the published version’s timeline with no corresponding content in the raw recording at all shouldn’t happen if you’re comparing your actual unedited source — if it does, check whether you’re looking at the true raw take or an earlier edit pass that was already trimmed.

Distinguishing a cut from a natural pause

Silence in the spectrogram — a quiet band with low energy but no clean break — is a pause, not a cut. What marks an actual edit point is the discontinuity: the acoustic character on either side (room tone, background noise, even the speaker’s breathing pattern) doesn’t connect the way it would across a continuous recording. A pause preserves the room; a cut usually doesn’t, even a well-hidden one, because two different moments in a recording session rarely have identical background noise.

When you don’t have the exact raw recording

If you only have your own separately-recorded copy of the same conversation — a different microphone, a different device — rather than the same file the publisher edited from, expect the comparison to run lower overall since you’re comparing two independent captures of the same event, not two edits of the same source file. Focus on the spectrogram’s structural pattern rather than the raw similarity score in that case: does the sequence and content of what’s present match, even if the exact acoustic signature differs because it came from a different recording device.

Try it

Compare audio → — the spectrogram difference view is what actually answers this question; the similarity score alone will tell you that something changed but not where.