Home › Tools › Remove Background Noise

Remove Background Noise

Clean up a recorded voice with a neural network that runs on our own servers, then normalize the level to the podcast standard. Long files are accepted: up to three hours, with no editing-software subscription.

  • Secure processing, files wiped after 60 min
  • No sign-up required
  • Up to 512 MB

Drop your files here

Accepted formats : .mp3, .wav, .m4a, .aac, .ogg, .oga, .flac, .wma, .opus, .mp4, .mov, .mkv, .avi, .webm, .wmv, .flv, .m4v, .mpeg, .mpg
Up to 512 MB per file.

    Options

    Clean up a recorded voice with a neural network that runs on our own servers, then normalize the level to the podcast standard. Long files are accepted: up to three hours, with no editing-software subscription. The tool accepts MP3, WAV, M4A, AAC, OGG, OGA files and works straight from your browser: nothing to install, no account to create.

    Under the hood, the work is done by DeepFilterNet 3, a denoising neural network running on our own processor, followed by EBU R128 normalization. Processing runs on FFmpeg, the open-source reference in the field, with encoding settings chosen to stay playable everywhere.

    Your files are processed on our European servers and then wiped automatically: they are never read, never kept, and never reused to train anything.

    When is this tool useful?

    • Pull the audio track from a talk or a lecture
    • Slim down a video that is too heavy to send
    • Produce subtitles from a recording
    • Convert media to a format your player can read

    How to do it

    1. Drop your file in the area above, or click to pick it.
    2. Adjust the options to suit your needs.
    3. Click “Remove Background Noise” and wait while it processes.
    4. Download the result. The link stays valid for 60 minutes.

    Frequently asked questions

    How is this different from a standard noise reduction filter?

    A standard filter subtracts an estimated noise spectrum and damages the voice along with it. Here, a neural network tells speech apart from everything else. In our measurements, at a 5 dB signal-to-noise ratio, the standard filter gains 2.5 dB where this processing gains 6.9.

    Is there a length limit?

    Three hours per file, where most free services stop at thirty minutes. It is built for conferences, lectures and long interviews.

    Do you keep my files?

    No. They are deleted automatically from our servers shortly after processing, along with the output file and the download link.

    Do I need an account?

    No, the tool works right away. An account is only needed to lift the daily limits.

    Does the result carry a watermark?

    Never. We add no mark of any kind to the files we produce, including on the free plan.