# Clean Up Audio: Clean up audio and sound like a studio.

> Remove background noise from voice recordings and videos with on-device AI, then get warm, clear studio sound at a comfortable level. Preview it, then save. Nothing is uploaded.

On-device AI takes the fan, hum, traffic and keyboard out of your voice, then shapes it like a studio recording at one comfortable level. Works on audio and video, and you hear the before and after first.

Part of [OneMacApp](https://onemacapp.com), 43 native tools for macOS 26+, one-time, yours forever. Page: https://onemacapp.com/clean-up-audio

## How it works

1. **Drop audio or video**: Interviews, lectures, podcast takes, screen recordings, voice notes. MP3, M4A, WAV, AIFF, FLAC, OGG and Opus open, and so do MOV, MP4, MKV and WebM videos.
2. **Preview the whole thing**: Pick Light, Medium or Strong and press Preview Clean-Up. The entire recording is cleaned, then switch between Original and Cleaned anywhere while it plays, with the video in view.
3. **Press Clean Up**: Cleaned copies land in a dated OneMacApp folder in Downloads. A video stays a video, with its picture untouched. The originals stay as they were.

## Features

- **AI noise removal, on your Mac**: A speech enhancement model runs on your Mac’s Neural Engine and keeps the voice while taking out what isn’t: fans, hum, traffic, keyboard clicks and room noise. It works several times faster than real time.
- **Warm, clear studio sound**: Then your voice is shaped the way a studio would: rumble cut, boxiness eased, a touch of warmth and presence, harsh “s” sounds softened and gentle compression for an even, close-miked feel.
- **One comfortable level**: Even Out Volume measures the whole recording and sets it to −18 LUFS in one steady step, so there are no sudden jumps and pauses aren’t pumped up. Comfortable for long listening.
- **Video in, video out**: Clean the sound of a video without exporting it first. The picture is copied exactly as it was, so there’s no quality loss, and the sound stays in sync.
- **Compare anywhere**: Preview cleans the full recording with a percentage as it goes, then lets you flip between before and after at any moment, as often as you like.
- **Private by design**: Everything runs on your Mac, including the AI. Calls, interviews and lectures are never uploaded anywhere.

## Specs

- **Opens**: MP3, M4A, WAV, AIFF, FLAC, OGG, Opus, MOV, MP4, MKV, WebM
- **Saves as**: Audio: M4A, MP3, WAV or FLAC. Video: same format, picture untouched
- **Noise removal**: On-device AI, Light, Medium or Strong
- **Even Out Volume**: −18 LUFS for the whole file, peaks under −1.5 dB
- **Preview**: The whole recording, before and after
- **Needs**: FFmpeg, installed for you in one click

## Questions

### How do I remove background noise from a recording on Mac?

Drop the file into Clean Up Audio and press Preview Clean-Up to hear the whole recording cleaned. Start with Medium. Switch between Original and Cleaned to compare, then press Clean Up to save.

### Can I clean up the audio in a video?

Yes. Drop the video straight in. Only the sound is cleaned: the picture is copied exactly as it was, and you get a video back in the same format (or MKV for less common ones).

### Which noise reduction level should I pick?

Medium suits most recordings. Light keeps a little of the room so the voice sounds natural and unprocessed. Strong removes the background completely, for very noisy places.

### Is it really AI, and does it upload my audio?

It uses a speech enhancement AI model that runs on your Mac, on the Neural Engine. Nothing is uploaded, there’s no account, and it works offline.

### What does Even Out Volume do?

It measures the loudness of the whole recording and brings it to −18 LUFS in one steady step, a calm level that’s comfortable to listen to. Because it never rides the volume moment to moment, pauses and breaths aren’t pushed up.

### Can it remove music or other voices?

It keeps speech and takes out most other sound, so background music and noises usually go. Other people talking are speech too, so they mostly stay.
