How to Clean Up Interview Audio: A Practical Step-by-Step Guide

Interview audio is harder to clean than a solo recording because it arrives messy in predictable ways: two voices at different levels, room noise on both tracks, one guest louder than the other, and a hum from a light or a fan that neither of you noticed while talking. The good news is that interview cleanup follows a repeatable sequence. Work in the right order, make each step small, and a raw two-track recording can become something that sounds like it was captured in a treated room. This guide walks through the full process, from preparation to final export, with the specific thresholds and settings that keep the voices natural.

Why Interview Audio Is Different

In a solo podcast you control one microphone and one room. In an interview you inherit whatever the guest's side sounded like — often a phone call, a video chat, or a second microphone placed on a different desk with a different noise floor. The two tracks may differ in tone, in level, and in noise. The job is to make them sound like they belong in the same conversation, then to remove the distractions that pull a listener out of it.

That means working on each voice separately before you combine them, and treating the cleanup as a chain rather than a single magic fix.

Before You Edit: Gather What You Have

Collect everything: the host track, the guest track, any backup recording, and — most importantly — a stretch of room tone from each location. Room tone is ten to thirty seconds of silence recorded in the same room with the same microphones, and it is the reference many cleanup tools use to learn the noise they should subtract. If you have it, you are ahead. If not, you can often find a quiet gap between sentences to sample instead.

If the interview was recorded on separate devices, note whether they used the same sample rate. Mismatched rates cause drift and pitch problems when you combine them, so convert to a common rate — 48 kHz — before you align anything.

Step 1: Sync and Separate

If you have two separate files, align them so the voices line up in time. A clap at the start of the session, or any sharp shared sound, gives you a visual and audible marker to line up on. Zoom in to the waveform and match the peaks. Drift of even a few milliseconds is audible as a slapback echo when the tracks play together, so check the alignment again near the end of the recording, not just the start.

Once aligned, keep the voices on separate tracks for now. You will process each one independently.

Step 2: Trim and Remove Silence

Cut the dead air first. Long gaps, the shuffle before the interview begins, and the "ums" that stretch into silence all make a recording feel slow. Removing silence tightens the pacing and shortens the file, and it also gives later tools less empty noise to amplify.

Do not cut every breath — removing all breaths makes speech sound unnatural and rushed. Trim the long pauses, the false starts, and the administrative chatter. Leave the natural short gaps that keep conversation flowing. A simple approach is to remove silence longer than about 0.7 to 1.0 seconds, using a tool that can do it automatically, then review the cuts by ear.

Handling the Two Tracks Differently

The guest's track often has more noise than yours. Trim each independently; do not assume the same silence threshold works for both.

Step 3: Reduce Noise on Each Track

Now tackle the hum and hiss. Apply noise reduction to the host track and the guest track separately, using each one's own room tone as the sample.

Keep the reduction modest. Aggressive noise reduction leaves behind a watery, metallic artefact, sometimes called "musical noise," that is more distracting than the original hum. A reduction of around 6 to 12 dB on the noise floor is usually enough for a steady hum; the ear forgives a faint background over the sound of a robot. Apply it in a couple of light passes rather than one heavy pass, checking after each.

For a constant low hum from electrical equipment, a high-pass filter set between 80 and 120 Hz removes much of it without touching the voice. For broadband hiss, the noise-reduction tool is the right choice.

Separating Voice From Music

If the interview sits over a music bed, or if you only have a mixed file where the voice and background music are combined, you can separate the voice from the music to isolate the speech. A stem-splitting tool does this by analysing the frequencies and stereo image that music and voice occupy. It is never perfect — separation always leaves a little artefact — but it can rescue an interview recorded in a noisy café or one where music was playing.

Step 4: Match Tone and Level Between Voices

With both tracks clean, make them sound like one conversation. Two voices from different rooms and microphones differ in tone and level, and the mismatch is jarring.

1. Apply the same high-pass filter (80 to 100 Hz) to both, so both lose low rumble equally.
2. Use gentle EQ to match brightness. If one voice is duller, a small boost of 1 to 3 dB somewhere between 2 and 5 kHz can add presence. If one is harsh, a small cut in the same range smooths it. Move in small increments; 1 to 2 dB is often all it takes.
3. Level-match the two tracks so both voices average similar loudness. Set your host and guest to the same target, such as −18 dBFS average, before the final loudness stage.

Do the tonal matching by ear, switching back and forth between the two voices, aiming for "same room" rather than "identical."

Gain and Compression

After matching, apply gentle compression to each voice so quiet words and loud laughs sit closer together. A ratio of around 2:1 to 3:1, a threshold set so only the louder parts are caught, and a soft knee will do the job. Too much compression flattens the life out of a conversation, so aim for consistency, not for maximum loudness.

Step 5: Assemble and Finalise

Now combine the two cleaned, matched tracks into your final mix.

1. Pan two voices subtly if you are working in stereo — for example, host slightly left and guest slightly right — but keep it gentle; hard panning is disorienting.
2. Check the whole conversation for level jumps where the two tracks meet, and smooth any that stand out.
3. Match the overall loudness to your delivery target — around −16 LUFS for a stereo podcast or −19 LUFS for mono — with true peaks no higher than −1 dBTP.
4. Listen end to end on both headphones and a small speaker, because problems hide in one and show up in the other.

If you want a faster path for the early steps, a web tool can remove silence and clean background noise from each track before you match and finalise. Offex Audio runs those steps on our secure server, which suits interviews recorded on a laptop when you would rather not install a full editor just to trim silence and tame a hum.

A Quick Troubleshooting Guide

• Voice sounds robotic after cleanup: you over-applied noise reduction. Undo and use less, in lighter passes.
• One voice sounds distant: it was recorded farther from its microphone. Get closer in future sessions; for now, a gentle presence boost and compression help but cannot fully fix distance.
• Echo when both tracks play together: the tracks are misaligned by a few milliseconds. Re-check your sync at the start and end.
• Sibilance too sharp after EQ: add a light de-esser or reduce your high-frequency boost.
• Loudness jumps between speakers: revisit the level matching in step four; the two tracks were not brought to the same average.

Conclusion

Cleaning interview audio is a sequence, and the sequence is what makes it work. Gather your tracks and room tone, sync and separate, trim the long silences, reduce noise on each voice using its own sample, match the tone and level between speakers, then assemble and set your final loudness. Keep every step gentle and check constantly, because the goal is not a sterile recording — it is two voices that sound like they shared the same room and the same conversation. Follow the order, resist the urge to over-process, and even a rough two-track interview becomes something a listener can enjoy without noticing the technical seams at all.

Frequently asked questions

Try the free audio tools →

Related guides and tools