---
title: "Noise Remover: Clean Up Speech on Your Device with AI"
description: "Reduce steady background noise in speech using RNNoise, an AI model that runs in your browser. Your recording is never uploaded."
canonical: https://toolsugar.com/tools/noise-remover
image: https://toolsugar.com/og.png
---

# Noise Remover

ToolSugar cleans background noise from speech with RNNoise, a small neural network running in your browser through WebAssembly. No audio leaves your device. On our test file the noise floor fell about 45 dB while speech level moved 0.2 dB. It is made for voices, not music. The output is mono. The first run downloads a 32 MB engine and a 5 MB model.

Runs on your device

## How it works

1. Choose a recording or a video with speech.
2. For audio, pick MP3, M4A or WAV.
3. Press remove noise and wait while it processes locally.
4. Listen to the result, then download.

## Key facts

- **Model**: RNNoise, an open recurrent neural network built for speech, running on your device.
- **Measured result**: Noise floor from -38.7 dB to -84.3 dB, speech level from -17.5 dB to -17.7 dB.
- **Test file**: A synthetic voice mixed with pink noise. Real recordings will vary, and we have not measured other kinds of noise.
- **Output**: Mono at 48 kHz. Stereo recordings are mixed down first.
- **Video**: Saved as MP4. H.264 and HEVC pictures are copied untouched, and VP8 or VP9 pictures are re-encoded to H.264.

## Common uses

- Cleaning up a voice note with steady noise behind it.
- Reducing background noise in a video call recording.
- Tidying a talk recorded with room noise, then checking the result by ear.

## Questions

### Is my recording sent to an AI service?

No. The model runs in your browser through WebAssembly. Our tests check that the page only contacts this site and the engine CDN while the tool runs.

### Will it work on music?

No. RNNoise is trained to keep speech. A steady two-tone test file dropped about 57 dB, so expect it to remove or damage instruments.

### Can I set how strong it is?

No. RNNoise makes its own decision for each 10 ms of audio, and there is no strength control.

[Read the Noise Remover guide](https://toolsugar.com/blog/how-to-remove-background-noise-from-audio)

## Related tools

- [Increase Volume](https://toolsugar.com/tools/increase-video-volume)
- [Audio Cutter](https://toolsugar.com/tools/audio-cutter)
- [WhatsApp Audio to MP3](https://toolsugar.com/tools/whatsapp-audio-to-mp3)
- [Extract Audio](https://toolsugar.com/tools/audio-extractor)