Remove Background Noise from Audio
Strip steady and shifting background noise — air conditioning, fans, street hum, room tone — out of any recording while keeping the voice natural. Built on DeepFilterNet, a full-band 48 kHz speech denoiser.
How it works
Your file is decoded to 48 kHz and passed through a deep filtering network that estimates a per-frequency-band gain for every short frame, suppressing noise energy while preserving the speech envelope. No noise profile or manual selection is needed.
What it's good for
- Podcast and interview recordings
- Zoom / Meet / Teams calls
- YouTube voiceovers
- Field and phone recordings
Details
- Engine
- DeepFilterNet
- Formats
- MP3, WAV, M4A, FLAC, OGG, AAC, MP4, MOV
- Price
- Free to try
Frequently asked questions
Yes. Unlike Audacity's noise-reduction, the model learned what speech and noise look like from thousands of hours of audio, so it removes noise automatically with no noise-profile step.
No. The denoiser applies a soft per-band gain rather than gating whole frames, so the natural timbre and breaths of the voice are kept while the steady noise floor drops.
Free users get a short demo clip; signed-in users can process full-length files up to the plan's size limit. Very long files are streamed in chunks.
It strongly reduces non-speech noise; competing speech is partly suppressed but for a talking background you'll get cleaner results from the dedicated crowd-noise tool.
Yes. MP4 and MOV uploads are accepted; we extract the audio track, denoise it, and return the file so you do not have to demux it yourself.
Yes. DeepFilterNet operates on the acoustic shape of speech versus noise, not on words, so it cleans any language equally well.
Your upload is used only to produce the cleaned result and is removed from our servers automatically afterward; it is not used to train the model.
It is tuned for voice, so on music it can dull instruments because it treats anything non-speech as noise. Keep it for spoken-word material and use the separation tools for music.