Updated October 2026 · Cleaning speech recordings
A video call recording is compressed and often uneven, but it can be cleaned for publishing. Record locally where possible, balance each voice, reduce steady noise and echo gently, and check the result on a phone. Expect limits: a call recording will not match a studio file, so set the expectations of the audience.
Video calls are now a common source of interviews, panels and conversations that people want to publish. The recording tool is convenient, but the files are compressed, the voices can sound uneven and background noise is often mixed in. A call recording can still be cleaned into a listenable episode, provided you start with the best file available and keep the processing honest.
This post explains how to get the best file, how to balance and clean the voices, how to reduce the echo and noise without damaging the speech and what limits to expect when publishing a call.
The cloud recording of a call is a compromise. Many calling tools also allow each person to record locally, which gives a cleaner file with no network compression. Ask each participant to record their own side if the tool allows it, and to keep the recording running for the whole call. Local files can then be aligned later, which gives you a much better starting point than the shared file alone.
Ask participants to use headphones, to sit in a quiet room and to keep a steady distance from the microphone. Turn on any noise setting the tool offers, but test it first, because it can make the voice watery. Check the connection, since a weak link causes dropouts that are very hard to repair.
Call recordings often have one loud and one quiet speaker. Level each voice separately, using gain changes on each track if you have separate files. Aim for similar loudness across the speakers, and keep the natural variation within each voice. The levelling guide covers the method in detail.
If a participant hears their own voice through speakers, the recording may carry an echo. Asking that participant to use headphones is the best fix. Noise reduction should be light and used only on the parts that need it. Check the speech on headphones, and make sure the consonants still sound clear.
Dropouts, frozen audio and sudden jumps in level all come from the connection. Mark each glitch as you listen, and decide whether to cut, repair or re-record a short section. A glitch that repeats a word can often be removed by cutting, but a dropout that removes a word may need the speaker to repeat the line.
Listen to the finished episode on headphones and on a phone speaker. Check that the voices are balanced, the noise is controlled and the speech is clear at a normal listening level. Check the loudness of the export with the LUFS meter, so the episode sits comfortably beside your other shows.
A call recording has a particular sound, and listeners may notice it. Tell them the episode was recorded on a call, so they know what to expect. That honesty builds trust, and it sets the right expectations for the sound.
When you have local files from each participant, treat them as the master and the cloud file as a reference. Compare the two at the start of the episode, and use the local tracks for the final edit wherever they are cleaner. The cloud file remains useful for checking what was said, and for filling any gap where a local file failed. Keep both, and name them clearly, so the choice is always visible.
Aligning the tracks is a small but important step. Clap at the start of the recording, or ask each speaker to say a short phrase, then line the tracks up on that moment before you edit anything.
Test the full chain before the real call. Record a short conversation with the same people, using the same apps and connections, and listen to both local files and the shared recording. Problems with a microphone, a connection or a setting are far easier to fix on a test than in the middle of an important interview.
A call recording will never sound exactly like a studio recording. The compression, the network and the home rooms all limit the result. The aim is a clear, honest episode that listeners can follow without effort. Pushing the processing too far makes the voices sound odd, so keep the work light.
The best starting point is a local recording. The Noise-Free Voice Recorder can clean an existing recording, and the Tunedzy Go page covers cleaning phone recordings. For recording an interview on a call, read the phone call post in this series.
Record voice-overs with live noise removal, or clean an existing file. Runs in your browser, nothing is uploaded.
Yes, with the right consent and a clean edit. Tell listeners it was recorded on a call.
Often it is a compromise. A local recording from each person is usually better.
Echo from speakers and network dropouts. Headphones and a stable connection prevent most of both.
Test it first, and keep it light, because it can make the voice watery.
