Pr Premiere Pro macOS

Hours of multi-mic audio,
cut in a minute.

Podcasts, interviews, panels, roundtables — anything where several microphones are running at once. Hush works out who is actually speaking, from how far each voice travelled rather than how loud it is, then mutes every other microphone and fades each edit.

Premiere Pro 2022 or later Β· macOS Β· cancel any time

Get Hush Try it free The free version runs on the first 5 minutes of any sequence.

One mic open at a time.

Five people around a table means five microphones hearing five people. Hush reads every track, decides who has the floor moment by moment, and disables the rest β€” so you keep one clean voice instead of five copies of the same room.

Real segment data from a 112-minute episode. Green is live, dark is muted. Every edit gets a crossfade shaped from your attack and release settings.

It knows spill from speech.

When mics are close together, one person's voice arrives on everybody else's track only a few decibels down β€” too close for a level comparison to help. Hush also compares the shape of the signals: a mic that is only hearing someone else rises and falls with them exactly, and a second person genuinely talking does not.

A2 speaking βˆ’17.1 dB
A1 hearing A2 βˆ’27.0 dB
A1 speaking βˆ’21.2 dB
Measured on a real session: only 9.9 dB between a voice and its spill. On that episode, shape-based rejection removed 32 minutes of echo that level alone had let through.

Fast enough to use on every episode.

Hush rebuilds the sequence as a file and re-imports it, rather than making thousands of edits one at a time. The difference is not marginal: a full episode is done while you are still reading the log.

70 stwo hours across five microphones
8,900cuts, each with a fade
0changes to your original sequence

Built for real recordings.

What it did to a real session.

Measured against hand‑labelled ground truth — a person marking who was actually speaking, second by second — on a five‑microphone recording. Not a demo reel: the same numbers were used to decide whether each change to the detection was worth keeping.

Clean isolation
one mic open, not several
before 19% after 64%
Leakage
a mic open over someone else
before 57.8% after 15.1%
Two failure modes, never averaged into one score: words lost and leakage let through. Averaging them hides the one that ruins an edit. Hush is deliberately tuned to leave a mic open rather than risk clipping the start of a word — on this session that cost 10% of words being caught slightly late, and you can push it either way per microphone.

It listens for who is nearest.

Loudness alone cannot separate two people sitting close together — a quiet talker on a hot mic reads louder than a loud one further away. So Hush also measures when the sound arrived. Sound covers about a metre every three milliseconds, so the microphone nearest the speaker hears them fractionally first, and that ordering does not care how loud anyone is.

speaking A2 0.9 m A1 1.8 m arrival A2 first A1, 2.6 ms later
2.6 milliseconds is roughly 125 samples at 48 kHz — small, but far larger than the measurement error, and it does not move when someone leans back or raises their voice. Hush measures it continuously, so microphones that get picked up and moved mid‑session are followed rather than assumed.

Every rule is a slider.

No two rooms sound alike, so nothing here is fixed. Move a control and the whole edit changes shape. Try it — this is the real arithmetic, running on a sample of one of my own episodes.

0 clips 0 fades 13 / 188 ms fade in / out
One minute of a two‑host conversation. The waveform is what each mic actually picked up, spill included — only the lit clips survive. Bridge gaps merges pauses shorter than the value, so a breath mid‑sentence stops becoming two clips. Fade length is the average; release decides how much of it lands on the tail, because speech starts fast and ends slowly.

One file. Five microphones.

Field recorders write every mic into a single multichannel file, and Premiere lays them out as tracks that all point at the same media. Hush works out which channel belongs to which track and decodes them separately — otherwise every track is the same mixdown, and no amount of tuning can tell five identical signals apart.

REC‑0042.WAV 10 channels · 48 kHz
It also follows a recorder that rolls over to a second file mid‑session, so a cut after the changeover still reads from the right one.

Not a silence remover.

Tools in this space usually remove the quiet bits from one track, or switch camera angles based on who is loudest. Hush does something else: it decides, moment by moment, which microphone should be open, and disables the rest.

Watch it run.

Two minutes, on a real episode: picking the mic tracks, saying what kind of recording it is, and what the timeline looks like afterwards.

Demo video coming soon.

Try it on your own audio.

The free version is the full product, limited to the first five minutes of any sequence. Same detection, same cuts, same fades β€” so what you hear is exactly what you would get. Paste a licence key to unlock the full length.

Monthly Annual SAVE 17%
£228 £190 /year

Billed yearly · works out at £15.83/month

Get Hush

2 months free · save £38 compared to monthly

  • Unlimited sequences, unlimited length
  • One machine, movable whenever you like
  • Every update while subscribed
  • Cancel any time · VAT and sales tax included

Questions

What do I get when I buy?

An email with the installer and your licence key. Drag the file onto the free ZXP Installer, restart Premiere, paste the key into the panel. No Terminal, no command line.

How many computers?

One at a time, and you can move it whenever you like. Remove from this Mac in the panel frees the licence immediately, so switching between a laptop and a studio machine, or replacing a computer, never needs an email to support.

Does it need an internet connection?

Only to activate. After that Hush re-checks quietly whenever you are connected, and a dropped connection will never stop you mid‑edit. You can work on a train, on location, or with the wifi off.

Will it change my original sequence?

No. It works on a duplicate and leaves yours untouched. Cuts are muted rather than deleted, so nothing is ever lost.

One thing to know: because the new sequence is created by rebuilding it as a file, Premiere’s importer leaves its Color Setup on “Custom” with a blank output colour space, which can make the new sequence preview darker. Exports are unaffected — verified by comparing rendered frames. Setting Sequence Settings → Color Management → Color Setup to match your original fixes the preview, and Hush reminds you on the panel after every run.

Where does it work well, and where does it struggle?

Best: every microphone recorded into one device — a Zoom, Rodecaster, Sound Devices or similar multichannel recorder — or separate files that started at the same moment. That is where arrival time is available, and it makes a large difference.

Hardest: separate recorders started at different times, with the microphones very close together. On one such session a mic picked up its neighbour louder than its own owner, and the recorders were not sample‑locked, so neither loudness nor timing could separate them. Hush still improves it — around 8% of words caught late and 33% leakage — but it will not be transformative. That is physics rather than a bug, and no setting fixes it.

The free 5‑minute mode exists precisely so you can find out which of these your recordings are before paying anything.

Premiere shows a β€œTranslation Report” when it finishes. Is that an error?

No. Hush rebuilds your sequence as a file and imports it back, and Premiere raises that notice on any imported sequence β€” it is telling you a report was written, not that anything failed. Click OK. A sequence containing nests imports as more than one sequence, so you may see several.

Does it need a multichannel recorder?

No, but it does more with one. Judging by mic distance needs every microphone to be a channel of a single recording, because that is the only way their clocks match β€” separate recorders drift apart by far more than the millisecond being measured. With a file per microphone Hush says so in the log and decides on level instead, which is how every other tool works. It is at its best on a field recorder.

Windows?

Not yet. Hush is macOS only for now.

Refunds?

Fourteen days, no questions asked — email hushextension@gmail.com and it will be refunded and cancelled. The free version runs the full detection on your own audio first, so you should know whether Hush suits your recordings before you spend anything.