Audacity Truncate Silence: Best Settings for Voice and Podcasts

What are the best Truncate Silence settings for voice?

The settings above shorten every gap of 0.8 seconds or more down to 0.4 seconds and leave everything shorter alone. That removes the dead air where someone checked notes or lost their train of thought, while the small pauses between clauses, the ones that make speech sound like a person talking, stay where they are.

The numbers are not arbitrary. They mirror AutoSilenceCut's Natural preset, which only touches gaps of 0.8 seconds or longer and keeps 0.32 seconds of each one. In Audacity I round the kept part up to 0.4 seconds because Truncate Silence cuts from the middle of the gap with no preview, and a slightly longer pause is easier to forgive than a clipped breath.

Where is Truncate Silence in Audacity?

1. Select the audio. Click in the track and press Cmd+A (Mac) or Ctrl+A (Windows) to select everything, or drag over one section.
2. Open the Effect menu, then Special, then Truncate Silence. This is the path in the 3.x series and in the current online manual, which was updated in October 2026 after Audacity 4.0 shipped.
3. If the item is missing, open the Plugin Manager from the Effect menu and check that the built-in effect is enabled. Audacity 4.0 launched without some Audacity 3 features, and the 3.7.x series (3.7.9 is the last patch) is still available if you need it.
4. Set the controls, click Apply, and listen. Cmd+Z / Ctrl+Z undoes it.

What does each Truncate Silence control do?

The dialog has two groups: Detect Silence and Action. Here is what each control means in practice, using the ranges and defaults from the official manual.

• Threshold (dB), range -80 to -20 dB, default -20 dB. Audio that stays below this level counts as silence. If too few gaps are caught, raise it (less negative). If words get cut, lower it.
• Duration (seconds), minimum 0.001, default 0.5. The audio must stay below the threshold for at least this long to count as a pause. This is the setting that protects natural clause pauses.
• Truncate Detected Silence with Truncate to (default 0.5 s). Every detected pause becomes exactly this long, whether it was 1 second or 6.
• Compress Excess Silence with Compress to (0 to 99.9 percent, default 50). Only the part of each pause beyond Duration is shrunk to this percentage, so long pauses stay longer than short ones.
• Truncate tracks independently (labelled Compress tracks independently in Compress mode), off by default. Off keeps all selected tracks in sync; on processes each track on its own.

The older manual adds a detail worth knowing: the effect shortens a pause by deleting from the middle of it. So with Truncate to 0.4 seconds, you keep about 0.2 seconds after the last word and 0.2 seconds before the next one. Empty space between clips counts as absolute silence and is always removed if it is at least as long as Duration.

How do you find the right threshold for your recording?

Every room is different, so measure instead of guessing. This takes about a minute.

1. Find a pause of a second or more where nobody speaks, and select only that pause.
2. Open Effect > Volume and Compression > Amplify. Do not apply it. The suggested Amplification value tells you how far the loudest sample in your selection sits below 0 dB. If it suggests 52 dB, your pause peaks at about -52 dB.
3. Cancel the dialog. Add 6 to 10 dB to that number: a -52 dB pause gives you a threshold between -46 and -42 dB.
4. Check your quietest speech the same way. Select a soft word ending and open Amplify again. Your threshold must sit below that level, or the effect will start eating words.

If the two numbers are within about 6 dB of each other, the recording is too noisy for any threshold to work cleanly. Reduce the noise first (our background noise guide covers the order of operations), or accept a longer Duration so only obvious gaps are touched.

Recommended settings by recording situation

The audiobook row uses a long Duration and Compress on purpose. Narrators use pause length to mark paragraph and chapter breaks, and Compress keeps a 3-second scene break noticeably longer than a 1.6-second sentence pause. Check your distributor's room-tone rules before shortening anything at the head or tail of a chapter; the loudness targets guide lists the ACX requirements.

Truncate or Compress: which action sounds more natural?

Truncate is predictable and easy to reason about: every long pause ends up the same length. That works for interviews and tutorials, where the long pauses are mostly hesitation. The downside is that a deliberate dramatic pause gets the same treatment as a fumble.

Compress preserves the shape of the speech. With Duration 0.8 seconds and Compress to 25 percent, a 3-second pause becomes 0.8 + (2.2 × 0.25) = 1.35 seconds, and a 1.2-second pause becomes 0.8 + (0.4 × 0.25) = 0.9 seconds. Longer still means longer. I reach for Compress on storytelling and narration, and Truncate on everything else.

Why do my podcast tracks drift out of sync?

Almost always because "Truncate tracks independently" was ticked. With it off, Audacity treats the selected tracks as one mix: time is removed only where every selected track is below the threshold, and the same edit is made to all of them. Your host's pause while the guest is talking is not a pause in the conversation, so it is left alone.

With it on, each track is processed separately. The host track loses its pauses, the guest track loses different ones, and within a minute the replies come before the questions. The manual suggests the independent mode for things like several unrelated songs in separate tracks, never for microphones on the same conversation. Also select all the tracks before applying, not just one, or only the selected track gets shorter.

One more trap from the 3.x manual: Truncate Silence does not act on gaps between separate clips unless "Editing a clip can move other clips" is turned on in the Tracks Behaviors preferences.

Common Truncate Silence mistakes

• Keeping the -20 dB default. It is the most common reason people say the effect "ruined" their voice. Measure and set the threshold properly.
• Duration below 0.5 seconds on conversation. Short clause pauses get flattened and the speaker sounds breathless.
• Truncate to 0 seconds. Words butt into each other and breaths vanish. Keep at least 0.1 seconds even for fast social edits.
• Running it twice. A second pass shortens the already shortened pauses. Undo and change the settings instead.
• Truncating before noise reduction in a noisy file. The threshold has nothing clean to grip. Clean first, then cut, then normalize.
• Forgetting labels and markers. The effect shortens the selection, so anything aligned to the old timing, such as chapter labels, needs checking afterwards.

When is a reviewable silence tool a better choice?

Truncate Silence is blunt by design. You choose settings, press Apply, and every gap in the selection is changed at once. There is no list of proposed cuts and no way to keep pause number 23 because the guest was thinking before an important answer, short of undoing and editing that region by hand.

The audio silence remover runs in the browser and shows every proposed cut on the waveform with Hear Before, Preview Cut and Restore controls, so you can keep the pauses that carry meaning. It processes the file locally; nothing is uploaded. You can stay in Audacity for the rest of the edit: choose "Audacity / Reaper Labels (.txt)" in the export format menu and import the file with File > Import > Labels. Each label spans a detected quiet gap and is named "Cut Silence" or "Restored Silence" with its number and length, so you can see exactly what the detector found before you cut anything.

For more on choosing values, see how to choose silence detection settings and why silence removal sounds choppy. Video editors can use the DaVinci Resolve and Premiere Pro guides.