Noise gate settings for voice

When the gate opens, the sound passes through, including background noise recorded alongside the voice. When it closes, the sound is reduced by the amount you set. Some tools mute the closed sections; others let you leave a quieter background. That second approach can sound less abrupt than switching between room noise and complete silence.

The gate responds to level, not to the meaning of a sound. A loud background sound can open it. A softly spoken answer can fall below its threshold. While someone speaks above the threshold, the background recorded with that voice can still pass through.

Use a gate when the main problem is unwanted sound in the gaps. If hiss masks words while a person talks, a noise-reduction trial addresses a different problem. If the voice itself is blurred by reflections, quietening the pauses does not remove those reflections from the words. Our guides to reducing audio hiss and reducing echo in a recording explain those separate limits.

Check the manual for your particular tool. Audacity calls its closing-time control Decay and uses lookahead in its gate. OBS uses separate open and close thresholds. A preset or a number copied from another recording cannot tell you where your quietest useful speech sits.

Make a working copy. Select a passage containing background alone, a normal phrase, the quietest answer you need to retain, and the pause before the next speaker. Include soft word beginnings and fading word endings. If speakers have different levels, include the quieter speaker too.

Listen before changing anything and note the exact words and sounds you need to preserve. For an important conversation, do not assume a breath, hesitation or quiet response is disposable background.

Measure or listen to the background-only part. Audacity’s Analyze Noise Level function can suggest a threshold from such a selection. Then compare that level with the quietest wanted speech.

On a negative-dB scale, −30 dB is a higher level than −40 dB. Moving a threshold toward zero raises the level needed to open the gate; moving it farther below zero lowers that trigger level. These numbers explain the direction, not recommended settings for your file.

Adjust the threshold so the gate can reduce the gaps while allowing that speech through. If a quiet answer is lost, lower the threshold and test again. If the background and quiet voice are too close in level for a useful setting, stop trying to solve that overlap with the gate. Raising the threshold further risks losing more speech.

Start with a modest amount of reduction rather than making every gap silent. Increase it only if the quieter gaps help. Listen across the join from a pause into speech: the background should not seem to jump in and out around every phrase.

A downward expander is another option when available. It reduces quieter material more gradually rather than using a hard cutoff. It still works from level; it does not identify which sound is a voice. Compare it with the gate on the same passage instead of assuming the gentler label guarantees a better result.

If the first sound of a word is missing, revisit the threshold and opening behavior. A faster attack or suitable lookahead may help, depending on the tool. Listen to the actual consonant; a moving meter does not prove it survived.

If the gate flickers open and closed inside a phrase, try more hold or adjust the difference between open and close thresholds where the tool provides it. This repeated switching is often called chatter.

If an ending is cut short, allow more hold or a longer release, then replay the whole phrase. The closing transition should preserve the natural decay of the voice. If the background audibly swells and drops, reduce the amount of attenuation and reconsider the timing. This rise and fall is often called pumping. Change one control at a time so you can hear what caused the difference.

Switch between the original and the trial at a similar listening level. Check the quiet answer, the beginning and end of each phrase, the gap, and the next speaker’s entrance. Listen once for intelligibility and again for changes to the background.

Keep the trial only if it helps without removing wanted material. Check other quiet sections and both speakers before applying the setting more widely. If a single setting cannot preserve them, use a more selective approach or leave the gate off. Save the processed version separately and keep a note of the tool and settings used.

An unusually silent gap can make a recording seem cleaner without making the words any clearer. Judge the result by the speech you can still hear, not by a flat-looking waveform between phrases.

For a recording that matters to a case or an important decision, keep the source and any listening copy separate. Do not remove pauses or quiet material from the only copy. If you need help assessing the recording, request an audio review and explain which passage you need to understand. An audio review can help assess the recording; it cannot guarantee that missing words can be recovered.

Audacity Noise Gate manual: threshold, reduction, lookahead, hold, decay and short-selection preview workflow.

Apple Logic Pro Noise Gate guide: reduction, timing, hysteresis and the distinction between a detector filter and the gated audio.

OBS Noise Gate Filter: separate open/close thresholds and attack, hold and release controls.

OBS Expander Filter: downward expansion and its adjustable reduction ratio.

Audacity Normalize manual: the direction of negative digital level values. This reference explains the scale; normalization is not a step in the gate trial.