Voice Editing

How to Set Recording Levels for a Voice That Gets Loud

Set recording headroom with a loud rehearsal, then find the first stage that clips before changing final playback volume.

Ask the speaker to read the loudest line in the script before setting the input gain. Ask for a laugh or an emphasized line if either will be in the take. A plain conversational reading is quieter than the passage you need to test. Keep the speaking distance close to the real take while you set the levels.

Watch the interface and the software meter during the test. Focusrite's input-gain guidance explains how loud sources and gain affect clipping and recommends leaving headroom. In its recording FAQ, Audacity suggests aiming for a peak around -6 dB on its recording meter. Treat that as a starting point for an Audacity test, and leave more room if the rehearsal was unusually controlled.

Find where the distortion begins

Record a 20-second rehearsal containing a normal voice, a breath or quiet ending, and the loudest material you expect to keep. Listen back with the playback level set comfortably. If the interface marks overload, lower its input gain and try the same phrase. Avoid adjusting several controls together. If the software meter peaks but the hardware is safe, check the selected input, any digital boosts or processors and the recording path. A lower monitoring volume can make distortion quieter without changing what went into the file.

The waveform is another clue. A repeated flat top close to the full digital scale can indicate clipping, but listen as well. Distortion can occur upstream, so a software meter below full scale may still record a distorted microphone, preamp or earlier processor. Change the distance or gain at the affected stage and make another recording while you can repeat the take.

Record a test before the real session

Note the microphone distance and interface setting after the rehearsal. If the speaker's louder line still has a margin and the quieter line stays intelligible, keep those conditions. Move the speaker away from the microphone for an especially loud passage if that keeps the level in range and the new room sound is acceptable. When multiple people share a session, set each input separately.

Listen to the file through a second output at a comfortable volume when the recording sounds distorted despite clean input meters. The monitor path can overload independently. Save both rehearsals and judge the version that plays cleanly through a known working output.

Save the unprocessed take before editing. Loudness normalization happens after capture, and turning a clipped file down cannot reconstruct missing waveform peaks. With a clean recording that has room for an unexpected laugh, you can decide how loud the finished file should play after editing.

Make the loud rehearsal sound like the session

Read the surrounding line as well as the loud word. An isolated shout after twenty seconds of silence may be delivered differently from a laugh during a lively exchange. For an interview, have the guest tell a short story or read the likely opening. Include a quiet ending and a naturally emphatic phrase. Keep their distance from the microphone steady while you find a usable gain. If the session calls for guests to turn toward each other, include that movement in the test.

Begin with gain low enough for the sudden sound and bring it up while watching and listening to that recording. Check the indicator available on the actual device. A color or scale on one interface may not mean the same thing on another. If your software has a peak hold, use it to recheck the line after the speaker finishes. Save the rehearsal and write down which phrase produced the highest level. If the loud moment happens only once, the recording gives you something you can inspect again.

For a worked example, the ordinary reading peaks at about -18 dBFS and the guest's laugh peaks near -4 dBFS. Raising gain to lift the everyday reading by six decibels would put that example laugh above full scale. Give the speaker room or leave the gain where it comfortably captures the loud part. Check the quiet voice on playback. If it is buried in steady room noise, improving the source or mic position can be more useful than increasing gain and risking the same overload.

Compare the meters in signal order

Use one close, harmless recording to check which changes affect the file. Start with the microphone and any input devices, then the record track, and finally the monitoring path. When the interface indicates an overload but the track meter does not, a control later in the signal chain may have turned down an already damaged signal. Lower gain at the input or move the source a little farther from the mic and test again. Check the new recording. An output fader does not protect an earlier input stage.

Listen to the same clean reference through your monitoring setup. If distortion follows a particular headphone or speaker output but disappears through another known output, inspect its level and connection before adjusting the input. If all playback paths reproduce the same rough sound on the rehearsal file, check whether the speaker or room actually produced that noise. Bypass a processor on a copy if it might be changing the signal. Test one adjustment and keep a record of which file came from it.

If two guests are recorded, try both voices talking at once. Set the inputs separately and check the mixed channel after routing. A mixed output can have less headroom than either isolated recording. Keep clean separate tracks when possible so that changing the final mix does not require re-recording a guest.

Write down what a later edit cannot repair

Mark any take that overloads at the source and record a second version while the speaker is still available. Label the second version with the input setup and the loudest phrase that passed. Tell the speaker how much change in distance they practised, especially if they cannot see the meters. Keep a short clean sample of the same position for later listening comparisons or room-tone edits.

Use the rehearsal as your baseline for that speaking distance and microphone. Recheck levels after a guest moves or a new microphone is connected. If the source stays clean, leave headroom and handle the final level at delivery. The podcast loudness measurement shows where to check gain and true peak once the edited program is ready.

Sources