About audioenhancer

Updated October 7, 2026

audioenhancer is a small web tool for one job: reducing steady background noise, such as a fan, an air conditioner or a computer, in spoken videos and audio you have already recorded. It is built and run by an independent developer.

Why it exists

Recording a tutorial or a talk at home often means a fan or an air conditioner humming in the background. The usual fixes are to record again, to learn an audio editor, or to pay for a subscription. This tool does the one job: upload the video and get the same video back with the steady noise reduced. One file up to 5 minutes is free, with no account and no watermark.

How it works

• The noise reduction is done by DeepFilterNet 3, an open-source speech-enhancement model, running on our own server. Your audio is not sent to a third-party AI service.
• The video picture is copied exactly and only the audio track is replaced. The length and the start time of the sound stay the same, so the voice stays in sync with the picture.
• The sound is mixed to mono for cleaning and placed in both channels of the result. If cleaning makes the voice quieter, the result is raised back to the original loudness (by up to 12 dB) with a limiter, so it does not clip.
• Cleaning takes about a quarter of the recording's length once the upload has finished: about 80 seconds for a 5-minute video.

What we have measured

On three real recordings made near a running fan, the estimated noise floor dropped by 39 to 52 dB, the loudness stayed within 0.6 dB of the original, and automatic transcription errors did not rise. The numbers and the method are on the fan-noise page. We also checked 18 kinds of files, from iPhone HEVC and portrait videos to OBS MKV recordings, camcorder MTS files and ProRes, for picture, length and sync.

What it is not for

Music or singing, gusty wind, room echo and several people talking over each other need different tools. If your recording has those, see other noise-removal tools compared.

Your files

Uploads are deleted after processing (preview-only uploads within 24 hours), cleaned results after 72 hours, and files are never used to train models. The details are in the privacy policy.

Contact

Questions, a result that did not clean well, or a correction to our comparison page: email jjj596651@gmail.com. We usually reply within one or two days.