How to clean up interview audio

Collect the recordings that already exist before exporting or combining anything. These might include a dedicated recorder, camera audio, a local interview recording, a call recording, or separate participant tracks. Keep the file names and note which device or session produced each one. Preserve the original files and make a separate copy for editing.

The Oral History Association recommends preserving original recordings, making redundant copies, and keeping related contextual records. For an interview, that context can include the participants, session date, agreed access restrictions, and notes explaining how the recording was made. Keep sensitive details out of file names or sharing links when they are not needed.

Compare sources at the same difficult passage. Listen to the quieter speaker as well as the louder one, and check another point later in the interview. Choose on the evidence in those passages rather than assuming the largest file, newest recorder, or quietest background must contain the clearest words.

If separate speaker tracks are available, keep them separate while assessing the problem. Rev’s recording guidance explains why individual tracks make it easier to address one speaker’s level or other issues without applying the same treatment to everyone. A two-channel file does not establish that each channel contains a different speaker: listen and check how it was recorded.

Write down the original timestamp and describe what you hear before choosing a tool.

• Quiet speech: Turning up the recording may also make its background noise more obvious. Doug Boyd’s oral-history recording guidance explains this limitation of a weak recorded signal.
• Steady hiss or hum: A treatment intended for consistent background noise may be worth testing. It is not the same job as repairing a click or separating overlapping voices.
• Clicks, pops, or a brief intrusion: Mark the time and listen before and after it. Keep the surrounding speech in the test. If the interruption covers a word, check another source rather than treating a smoother repair as confirmation of that word.
• Room echo: Speech is mixed with reflections from the room. Compare another microphone or source before trying to suppress it further.
• Harsh or distorted speech: Check whether the distortion is already in the original. Lowering playback volume does not undo recorded clipping.
• Overlapping speakers or missing sound: First check other available tracks. Do not fill a gap with words that are not established by the recording.

These distinctions help you choose a test; they do not predict whether a particular interview can be restored. The existing guides to room echo and speech that remains unclear cover those problems in more detail.

Choose a passage containing the difficult words and enough surrounding speech to judge it. Duplicate it or work in a project that leaves the original file untouched. Keep an unprocessed version available for comparison.

Change one treatment at a time. Compare at similar listening volume so a louder version does not win simply because it is easier to hear. Listen through the words, sentence endings, pauses, and changes of speaker. If speech sounds thinner, interrupted, or less distinct, reduce the treatment or undo it.

For example, Audacity’s Noise Reduction effect requires a representative selection containing noise alone. A pause containing a quiet syllable or breath you want to preserve is not automatically a suitable noise profile. The manual advises using the lowest reduction that gives an acceptable result and checking the Residue preview, which plays what the effect would remove. Recognizable wanted speech in that residue is a warning that the treatment may be too strong.

The same manual says this effect is intended for relatively constant noise and may give unsatisfactory results when noise is loud, variable, or similar to the wanted sound. That is why a single universal preset is a poor basis for processing an entire interview.

Keep the version that preserves useful speech, even if some background remains. Then check the exported listening copy, including the beginning, middle, and end. Record which source, passages, and treatments you used so another editor can follow the work.

Sound cleanup and editorial editing need different decisions. Adjusting level to make an answer audible is different from shortening the answer, changing its order, or removing context that changes what the speaker appears to say.

The Associated Press permits subtle standard audio processing, including level adjustments, equalization, and reduction of extraneous noise, provided it does not conceal, remove, or otherwise alter the audio’s content. Its rules also restrict changes to newsmaker audio and the use of substituted sound. Those are AP’s standards; apply the editorial rules governing your own project rather than treating a software feature as permission to change an interview.

Keep notes of content edits separately from technical cleanup. Do not generate a missing word or splice a new statement together and present it as the original answer. If a passage cannot be understood reliably, retain that uncertainty or choose another supported excerpt.

Before sending interview material to an external tool or service, check the agreed consent, access restrictions, and handling requirements. The Oral History Association’s guidance treats these agreements and the integrity of a narrator’s perspective as part of using the recording responsibly. This guide does not establish the legal permissions for your recording or promise that a particular platform satisfies your project’s requirements.

Listen to the question, the full answer, and the surrounding original audio before approving an excerpt. Check the speaker, wording, and original timestamp. A transcript can help locate a passage, but its text must be checked against the recording.

Mark uncertain words rather than completing them from what seems likely. If cleanup makes a possible word seem clearer, compare it again with the original and ask whether the sound supports that reading. When a word carries an important factual distinction, seek another editorial check before relying on it.

The Oral History Association recommends contextualizing excerpts and retaining the integrity of the narrator’s perspective. Keep a reference to the full source so a later reviewer can examine the passage rather than relying only on the cleaned clip.

Use a brief test with the actual people and equipment before starting:

1. Choose a quiet location and listen for persistent room noise, clothing rustle, and interruptions.
2. Record both voices and play the saved test back through headphones. Check the file, not only the live sound in the room.
3. Watch recording levels and leave room for an unexpectedly loud answer. Monitoring still matters after the interview begins.
4. Confirm that any separate tracks and backup recording are actually being saved.
5. Record the agreed session context and keep the source files afterward.

For a remote or screen-recorded interview, make one short capture that includes both a spoken question and the sound you intend to record from the computer. Play the saved file back and confirm both are present. Keep microphone and system audio separate when the recording system supports it. Retain the original export when available rather than making a screen recording of it.

Check microphone placement and any wind, pop, or handling protection against the equipment’s own guidance. Listen to the recorded test in the conditions you will use. A prevention check changes what reaches the microphone; it cannot repair a conversation that has already been captured.

Boyd’s recording guidance supports headphone and level checks; the University of Minnesota’s interview guidance also highlights the room, notifications, and clothing noise. The aim is to catch a real problem early, without turning interview preparation into an equipment shopping exercise.

If an important passage remains difficult, gather the original file and any alternative tracks. Include the relevant timestamps, what you need to hear, what processing you have already tried, the intended use, and any deadline or handling restrictions.

Media Medic’s audio engineer listens before an invoice is sent. File review is free; cleanup is paid work. An assessment can establish what work is appropriate for the supplied material, without promising that every word can be recovered. Keep editorial approval of a quotation separate from the cleanup decision.

This guide compares published audio-processing instructions, interview-recording guidance, and editorial and oral-history standards. It applies those sources to the decisions involved in assessing an existing interview: source selection, a small processing test, preservation, and quote checking. It does not report a restoration experiment or a measured improvement in transcription accuracy. Software instructions describe their named tools; editorial and access requirements depend on the project.