How to strip long pauses from any audio recording
- Step 1Drop the recording — Drag your meeting, lecture, or memo onto the silence-stripper tool. Audio or video both work, so a Teams/Zoom recording or a phone voice memo is fine. Everything stays local — nothing uploads.
- Step 2Raise the threshold for room audio — Room-recorded audio has a high noise floor (HVAC, laptop fan, far-field reverb), so the default
-40 dBoften won't trigger. Start at-35 dBand, if nothing is cut, raise toward-30/-28 dBuntil the gaps register. Range-80…-10, step 1. - Step 3Set the minimum silence (s) — Min silence (s) is the cushion left at each cut. For lectures and meetings
0.5–0.8 skeeps the cadence natural. Drop to0.3 sif you want a very condensed study version. Range0.1…5, step 0.1. - Step 4Run the strip — FFmpeg applies
silenceremovewithdetection=peak: leading silence trimmed once, every internal and trailing pause tightened. It removes only time — content, words, and pacing within speech are untouched. - Step 5Spot-check the result — Jump to a couple of pauses (e.g. after a question, during a board-writing gap). If words run together unnaturally, raise Min silence. If long stretches remain, raise the Threshold toward
-30/-28 dB. - Step 6Download the condensed recording — The output downloads as
recording-tightened.extin your original format. Built-in-mic recordings often benefit from cleanup next — run ai-noise-reducer for hum/hiss, then loudness-normalizer for consistent level.
The two real controls
These are the only options the silence stripper exposes, verified against the tool's client component and the FFmpeg filter built in the processor.
| Control | FFmpeg parameter | Default | Range / step | Meaning for room recordings |
|---|---|---|---|---|
| Threshold (dB) | start_threshold + stop_threshold (detection=peak) | -40 dB | -80 to -10, step 1 | Usually needs raising for room audio — the default -40 dB sits below a typical HVAC/fan noise floor, so nothing would be cut |
| Min silence (s) | start_silence + stop_silence | 0.5 s | 0.1 to 5, step 0.1 | Cushion left at each cut. Keep 0.5–0.8 s for lectures/meetings; go lower for a tightly-condensed study cut |
Settings by recording type
Starting points for common non-podcast recordings. Room-captured audio almost always needs a higher threshold than studio audio.
| Recording type | Suggested threshold | Suggested min silence | Notes |
|---|---|---|---|
| Voice memo, phone held close | -40 to -38 dB | 0.5 s | Near-field, fairly clean; close to defaults |
| Meeting via laptop mic | -33 to -30 dB | 0.6 s | Far-field with fan/HVAC; raise threshold to clear the floor |
| Lecture in a large room | -30 to -28 dB | 0.6–0.8 s | Reverb + room tone; loose threshold, gentle cushion |
| Study-condense (max tightening) | -30 dB | 0.25–0.3 s | Most aggressive; expect a clipped, fast-paced result |
Pro-tier limits for long recordings
The tool requires the Pro tier or higher. Size and duration are separate caps; whichever is hit first applies.
| Tier | Max file size | Max duration | Files per batch |
|---|---|---|---|
| Pro (minimum) | 200 MB | 120 min | 10 |
| Pro-media | 100 GB | Unlimited | 100 |
| Developer | 100 GB | Unlimited | Unlimited |
Cookbook
Recipes for condensing meetings, lectures, and memos with the two-control silence stripper. Because these are room recordings, the recurring theme is: raise the threshold until the gaps actually register.
Condense a 90-minute lecture
A large-room lecture captured on a far-field mic. Raise the threshold above the room tone and keep a gentle cushion so the lecturer's emphasis survives.
Input: lecture-week7.m4a (91 min)
Settings: Threshold -29 dB · Min silence 0.7 s
FFmpeg applies:
silenceremove=start_periods=1:start_silence=0.7:start_threshold=-29dB:
stop_periods=-1:stop_silence=0.7:stop_threshold=-29dB:detection=peak
Output: lecture-week7-tightened.m4a (72 min)
-> ~21% removed; board-writing gaps gone, pacing intactStrip a screen-share-heavy meeting
A team meeting with long silent stretches while someone sets up a screen share. Laptop-mic audio needs the threshold up.
Try -40 dB first -> nothing cut (fan noise > -40 dB) Fix: Threshold -31 dB · Min silence 0.6 s Input: standup.wav (58 min) -> standup-tightened.wav (47 min) -> 19% removed; the setup silences are gone
Tighten a rambling voice memo
A close-held phone memo with thinking pauses. Near-field audio is clean, so defaults work well.
Input: idea-memo.m4a (12 min) Settings: Threshold -40 dB · Min silence 0.4 s Output: idea-memo-tightened.m4a (9 min 30 s) -> ~20% removed; thoughts run back-to-back
Max-condense a lecture for revision
A study cut where speed matters more than natural cadence. Push both controls aggressive.
Input: lecture-revision.m4a (88 min) Settings: Threshold -30 dB · Min silence 0.25 s Output: lecture-revision-tightened.m4a (64 min) -> ~27% removed; fast, slightly clipped, fine for review
Denoise the room, then strip the gaps
When constant HVAC noise blocks cuts, remove it first so the real gaps drop below the threshold.
Step 1 - ai-noise-reducer (/audio-tools/ai-noise-reducer): meeting.wav -> meeting-clean.wav (RNNoise removes HVAC hum) Step 2 - silence-stripper: meeting-clean.wav -> -40 dB now works (floor is gone) -> meeting-clean-tightened.wav, gaps cut cleanly
Edge cases and what actually happens
Default -40 dB cuts nothing on room audio
By designRoom recordings (laptop mic, HVAC, far-field reverb) usually have a noise floor above -40 dB, so no region reads as silent and the output equals the input. Raise the threshold toward -33/-30/-28 dB until the pauses register. This is the single most common surprise on meeting/lecture audio.
Words run together after a tight cut
By designA small Min silence (0.2–0.3 s) removes nearly all the gap, so sentence boundaries blur. That's the cost of maximum condensing. Raise Min silence to 0.5–0.8 s to restore natural spacing between thoughts.
Constant HVAC/fan noise blocks all cuts
By designIf background noise never drops below the threshold, nothing qualifies as silence. Either raise the threshold above the noise level, or remove the noise first with ai-noise-reducer so the genuine gaps fall below the threshold.
A speaker's quiet aside gets clipped
ExpectedFar-field audio means a turned-away or soft-spoken aside can dip below the threshold and be treated as silence. Lower the threshold magnitude (e.g. -30 → -35) to protect quiet speech, accepting that fewer gaps will be cut.
Reverberant room tail counts as content
ExpectedIn a large echoey room, the reverb tail after speech can sit above the threshold, so the 'gap' starts later than the words end and less is cut. This is correct level-based behaviour; a higher threshold cuts into the tail sooner if you want tighter results.
Recording exceeds the Pro 200 MB / 120 min cap
RejectedA multi-hour meeting or a long uncompressed WAV can exceed Pro's caps (200 MB size, 120 min duration). Split it with audio-splitter, strip each part, and rejoin — or use Pro-media (100 GB, unlimited duration).
Output format matches the input memo format
ExpectedThere is no format selector — an M4A memo comes out as M4A (re-encoded once via the AAC encoder to apply the cut), a WAV stays WAV. If you want a smaller shareable file, convert afterward, e.g. with m4a-to-mp3.
Multiple speakers, uneven distances
ExpectedIn a conference-table recording, near speakers are loud and far ones are quiet. A single threshold can clip the far speaker's gaps or miss them. Pick the threshold for the quietest speaker you care about, accepting fewer cuts overall.
Frequently asked questions
Why does the default setting cut nothing from my meeting recording?
Room recordings have a high noise floor (laptop fan, HVAC, far-field reverb) that usually sits above the default -40 dB, so FFmpeg sees no silence. Raise the Threshold toward -33/-30/-28 dB until the gaps register. If constant noise still blocks cuts, remove it first with ai-noise-reducer.
Does my meeting or lecture audio upload anywhere?
No. The tool runs FFmpeg 8.1 in WebAssembly inside your browser tab — confidential meeting recordings and internal lectures are processed locally and never uploaded. The tool requires the Pro tier (200 MB / 120 min per file).
What threshold and minimum-silence should I use for a lecture?
For a large-room lecture start at -30/-29 dB with 0.7 s minimum silence. A close-held voice memo is cleaner and works near the defaults (-40 dB, 0.5 s). A laptop-mic meeting usually needs about -31 dB. Always confirm by ear and raise the threshold if too little is cut.
How much can I shorten a recording?
Meetings and lectures typically carry 15–25% silence, so a tighten of that order is realistic. With an aggressive Min silence of 0.25–0.3 s and a higher threshold you can push toward the upper end, at the cost of a clipped, fast-paced feel.
Will it remove the actual content or words?
No — it removes only time in regions below the threshold. Spoken words sit above the threshold and are preserved. The only risk is a very quiet aside or a soft onset dipping below the threshold; protect those by lowering the threshold magnitude.
What format does the condensed recording come out as?
The same format you uploaded — there's no format selector. An M4A voice memo stays M4A, a WAV stays WAV. The file is named recording-tightened.ext. To shrink it for sharing, convert afterward, e.g. with m4a-to-mp3.
Should I denoise before stripping silence?
If constant background noise (HVAC, fan) is blocking cuts, yes — denoise first so the real gaps fall below the threshold, then strip. If the recording is fairly clean, just raise the threshold instead. Denoise with ai-noise-reducer.
Can I batch a semester of lectures at once?
The Pro tier allows up to 10 files per batch (Pro-media 100, Developer unlimited). They process locally with the same settings. If lectures were recorded in different rooms, run them individually so you can tune the threshold per room.
My 2-hour meeting was rejected — why?
Either the size cap (200 MB) or the duration cap (120 min) on the Pro tier. A long uncompressed WAV hits the size cap quickly. Split with audio-splitter and strip each part, or upgrade to Pro-media (100 GB / unlimited duration).
Can it even out a quiet far speaker vs. a loud near one?
No — the stripper only changes timing, never level. For uneven volumes around a conference table, run speech-leveler after stripping, then loudness-normalizer for a consistent final level.
Does it work on video meeting recordings?
Yes — the tool accepts video input, decodes the audio track, applies the strip, and writes audio out. If you specifically need the audio extracted from a video first, video-to-mp3 and video-to-wav do that directly.
Is there a faster way to clean up a recording end-to-end?
For voice content that also needs noise removal and leveling, podcast-master does silence handling, denoise, leveling, and loudness in one pass. Use the standalone silence stripper when you want manual control over how hard the pauses are cut.
Privacy first
Every JAD Audio tool runs entirely in your browser via FFmpeg (WebAssembly) and RNNoise. Your audio files never leave your device — verified by zero outbound network requests during processing.