Cleaning up a recording on your Mac, without uploading it
Adobe Podcast Enhance is free but caps files at 30 minutes and one hour a day, and every file goes to Adobe's servers. Cleanvoice costs $11 a month for 10 hours. Vocemo removes filler words on your own Mac for $15 a month, and nothing is uploaded.
- Adobe Podcast Enhance
- Free, 30 minutes a file, 1 hour a day, cloud only
- Cleanvoice
- $11 a month for 10 hours, or $11 for 5 hours as you go
- Auphonic
- Free for 2 hours a month, then $13 a month for 9 hours
- The local option
- Filler words cut on your Mac, $15 a month
Why would you not want to upload an interview recording?
Most audio cleanup happens in a browser tab. You drag a file in, wait, and download a tidier version. The file has left your Mac by then, and for a lot of recordings that is the part nobody thought about.
Think about what is actually on the tape. A journalist's interview names a source. A user research session has a customer describing an internal system by name. A therapy supervision recording is covered by a confidentiality agreement. A legal intake call is privileged. A doctor's dictation is health data. A board discussion is price sensitive until it is not.
None of that is unusual. It is the ordinary content of the recordings people most want to clean up, because those are the ones made in a bad room with a cheap microphone and a nervous speaker who says "um" a lot.
Uploading is not automatically wrong. It is a decision, and it is easier to make once you know what the local option actually covers. More on the general shape of that trade in running AI on your own Mac.
How do you remove um and uh from a recording on a Mac?
Filler word removal is a text problem wearing an audio costume. The tool transcribes the recording, finds the words you asked it to drop, then cuts the matching slice of audio and closes the gap. Everything depends on the transcript being right, and on the cut being made at a sensible point rather than mid breath.
There are three routes on a Mac.
- A transcript based editor. Descript is the best known. You get a document, you delete the word, the audio changes. Descript's Free plan gives 60 minutes of media a month. Hobbyist is $24 a month, or $16 a month billed annually, for 10 hours of media a month.
- A dedicated cleanup service. Cleanvoice does filler words, stutters, mouth sounds and dead air in one pass. It is a web upload. Its free trial is 30 minutes.
- A local tool. Vocemo transcribes the recording on the Mac, marks the filler words and writes a cleaned file. No file leaves the machine. It costs $15 a month, or $144 a year, which works out at $12 a month.
One warning that applies to all of them. Removing every single "um" makes a person sound strange. Speech has rhythm, and the fillers carry some of it. Cutting the long ones and the stumbles is usually enough. Listen back before you commit, because a run of tight cuts is noticeable in a way that no waveform tells you about.
How do you cut long silences from a recording?
Silence trimming is separate from filler removal, and it is the one job you can do well for free on a Mac.
- Audacity. Free, and it runs natively on Apple silicon. Select the track, then Effect, then Special, then Truncate Silence. Set the threshold and the maximum silence you want to keep. It is not clever, but it is fast and it is local.
- ffmpeg. The
silenceremovefilter does the same from the command line and is easy to run over a folder of files. - An editor with a strip silence command. Logic Pro, Reaper and Descript all have one, and they let you see the cuts before applying them.
Be careful with interviews recorded on one track. A pause that looks like silence is often one person thinking, and cutting it changes what the exchange sounds like. Aim to shorten silences to about half a second rather than remove them.
How do you reduce background noise without uploading the file?
This is where the cloud tools are genuinely ahead, and it is worth saying plainly.
Adobe Podcast Enhance rebuilds the voice rather than filtering the noise. It takes a recording made in a kitchen and returns something that sounds like a treated room. Nothing running locally on a Mac matches it for that particular trick. It is free, on the free plan the file must be 30 minutes or shorter and up to 500 MB, and you get 1 hour of processing a day. Auphonic does the more conservative version of the job: levelling, loudness targets, noise and reverb reduction, and it is free for 2 hours of processed audio a month.
Locally you have three honest options. macOS has a Voice Isolation microphone mode, in Control Centre while an app is using the microphone, which helps at record time rather than afterwards. Audacity's Noise Reduction effect works well when the noise is steady, such as an air conditioner or a hum, and you have a couple of seconds of it alone to sample. And a vocal separation model can pull a voice away from music or a busy background: Vocemo does that on device, which is useful when the problem is a backing track rather than hiss.
If your recording is quiet, boxy and echoey, upload it or record it again. If it is a decent voice with steady noise or a music bed behind it, you can fix that on the Mac.
What do the audio cleanup tools cost?
Prices checked September 2026, taken from each vendor's own pricing page.
| Tool | Where the audio goes | Price |
|---|---|---|
| Adobe Podcast Enhance | Adobe's servers | Free plan: 30 minutes a file, up to 500 MB, 1 hour a day |
| Cleanvoice | Cleanvoice's servers | $11 a month for 10 hours, $30 for 30 hours, $90 for 100 hours. Pay as you go $11 for 5 hours |
| Descript | Descript's servers | Free: 60 minutes a month. Hobbyist $24 a month, or $16 a month billed annually, for 10 hours |
| Auphonic | Auphonic's servers | Free for 2 hours a month. Auphonic S $13 a month, or $11 a month billed yearly, for 9 hours |
| Audacity | Stays on your Mac | Free |
| Vocemo | Stays on your Mac | $15 a month, or $144 a year, which is $12 a month |
If you would rather this ran on your own Mac, that is what Vocemo does. Free for 7 days, then $15 a month for 3 Macs.
Download free for MacNote the shape of the pricing. The cloud tools charge by the hour of audio, so a heavy month costs more than a light one. A local tool charges by the month whatever you run through it, since the compute is your own Mac and it costs the vendor nothing.
What does the local option not do?
Worth being blunt about this, because the gaps are real.
- No speech rebuilding. There is no local equivalent of Adobe Podcast Enhance. A bad room stays a bad room.
- No automatic silence trimming in Vocemo. It cuts filler words. For dead air you want Audacity or an editor with a strip silence command.
- No loudness normalisation to broadcast targets. If you need a file at a specific LUFS figure for a podcast host, Auphonic does that and Vocemo does not.
- No multitrack levelling. Balancing four separate guest tracks against each other is an Auphonic and Descript job.
- No shared workspace. Nothing is on a server, so there is no link to send a producer and no comments thread.
- Apple silicon only, macOS 13 or later. An Intel Mac cannot run it.
What you get instead is that the recording never leaves the machine, the models come down only when a feature needs one and shut down after 120 seconds idle, and the installer is 5.5 MB.
Which tool for which recording?
- A confidential interview with steady background noise. Clean it locally. Filler words and a vocal separation pass, then trim silences in Audacity.
- A podcast episode recorded in a bad room, nothing sensitive in it. Adobe Podcast Enhance, free, then Auphonic for loudness if your host wants a target.
- Four remote guests on four tracks. Auphonic or Descript. Multitrack levelling is the thing they do that the local option does not.
- A recorded meeting you want written up. That is a transcript job more than a cleanup job. See meeting notes without a bot.
- Twenty episodes a month, none of them sensitive. Do the sums on hours. Cleanvoice at $90 a month for 100 hours is cheap per hour at that volume.
Questions
Can I remove filler words without uploading my audio?
Yes. A speech model running on an Apple silicon Mac can transcribe the recording, find the filler words and cut them without any network access. Vocemo does this for $15 a month. The trade is that a local tool cannot rebuild a badly recorded voice the way Adobe Podcast Enhance does.
Is Adobe Podcast Enhance free?
There is a free plan. On it a file must be 30 minutes or shorter and up to 500 MB, and you get 1 hour of processing a day. Files are processed on Adobe's servers, so it is not suitable for confidential recordings.
How much does Cleanvoice cost?
Cleanvoice subscriptions are $11 a month for 10 hours of processed audio, $30 a month for 30 hours and $90 a month for 100 hours. Pay as you go credits start at $11 for 5 hours. The free trial covers 30 minutes.
What is the best free way to remove silence from audio on a Mac?
Audacity, which is free and runs natively on Apple silicon. Use Effect, then Special, then Truncate Silence, and set the maximum silence you want to keep. Aim for about half a second rather than zero, or the conversation starts to sound clipped.
Does removing um and uh make a recording sound worse?
It can. Fillers carry part of the rhythm of speech, and cutting every one leaves a delivery that sounds unnaturally tight. Removing the long hesitations and false starts usually gets most of the benefit. Always listen back before you publish.
Can a Mac reduce background noise offline?
Partly. macOS has a Voice Isolation microphone mode that helps while you record. Audacity's Noise Reduction effect works on steady noise if you can sample a couple of seconds of it alone. A vocal separation model can lift a voice off a music bed. None of these match Adobe Podcast Enhance on a genuinely bad room.
Sources
Every price on this page was read from the vendor's own page on the date shown above. Prices change; if you find one out of date, tell us and we will correct it.
Competitor prices in this article were checked in September 2026. Prices change; if you find one out of date, write to us and we will correct it.