Voice Noise Remover (Free)

Reduce background noise in meeting, lecture and YouTube recordings. Switch between the original and the cleaned version and adjust the strength.

Before you start Copyright and recording rules (tap to read)

Loading the tool...

When you can't fully control where you record, an AI noise reduction tool can cut background noise such as wind or hum from a recording you already have.

When to use it

How to use

  1. Set the strength slider (0 to 100%) first. The default is 100%.
  2. Add your recording. mp3 · wav · m4a · mp4 and more are supported.
  3. The audio is converted to 48 kHz and the AI (RNNoise) reduces the noise. You'll see the progress and the share of the audio that contains speech.
  4. Compare with "Play original" and "Play cleaned", and adjust the strength while you listen.
  5. Save as WAV or MP3 (192 kbps).

How it works

The default engine is RNNoise compiled to WebAssembly and loaded from this site; the "High quality AI" option runs the GTCRN speech-enhancement model, a roughly 0.5 MB download the first time. Both run in your browser, so the recording is never uploaded or kept on a server.

Your files are processed in your browser and are not uploaded to a server. You can check this yourself.

Supported formats and limits

Tips

Troubleshooting

New
  • Choose the engine: Standard (RNNoise) or High quality AI (GTCRN).
  • A/B switching keeps the playback position.

For developers: Tools That Don't Upload offers a REST API with a free tier (usage limits apply) for data, text, code, PDF and web tasks. Audio processing is not part of the API.

FAQ

Can I use it on music?

We don't recommend it. The AI is tuned for human voices, so on music it can cut or distort instruments.

Will the recording be completely clean?

It reduces noise considerably but can't remove it entirely. Very loud noise, or other people talking in the background, is hard to remove.

How do I choose the strength?

Set it anywhere from 0 to 100% with the slider. Higher values remove more noise but can make the voice sound unnatural, so compare with the original to find the right value.

Does the file stay on a server after processing?

No. All processing happens in your browser and the file is never uploaded or stored on a server. The High quality AI option only downloads its small model (about 0.5 MB) the first time.

What does "High quality AI" do?

It runs GTCRN, a small speech-enhancement model (MIT license) served from this site, inside your browser. "Standard (fast)" uses RNNoise. Try both on your recording. Files never leave your device.

Can I clean the audio of a video and keep the video?

Yes. Drop an MP4 or MOV and, after cleaning, choose "Save as video (MP4)". The video track is copied without re-encoding, so the picture quality stays the same. WebM videos or browsers without WebCodecs AAC support can only save the audio (WAV/MP3). Latest Chrome or Edge is recommended.

Can I process several files at once?

Yes. Pick several files and they are cleaned one after another. Download each result or all of them as one ZIP. Audio files are saved as MP3, MP4/MOV videos as video.

Can I see how much noise was removed?

Open "Compare before and after on a spectrogram" under the result to see the original and the cleaned audio as frequency pictures. A darker background means less noise.

Can I send a recording straight from the Voice Recorder?

Yes. Press "Remove noise" in the recorder and the file opens here without picking it again. The file is only passed along inside this browser on your device. The handoff expires after 10 minutes; the file is removed when it opens here, or at the next handoff if it was never opened.

Learn more

For more detail, see the guides: How to Reduce Noise in Recordings (Before and After) and Vocal Remover: What Actually Works and Its Limits.

More tips

Tips for cleaner results

What it may not fully remove

Sudden loud events (a dog barking, a door slamming), music playing in the background and several people talking over each other may be reduced but can remain audible. If the noise is louder than the voice, the result can still be hard to understand.

Related tools