audioenhancer Clean a file

A real example + a tool to try

Remove fan noise from a video you already recorded

A fan or air conditioner can leave a steady hum under your voice. Reduce that noise in a talk, tutorial or demo while keeping the picture unchanged.

  • Up to 5 minutes free
  • No sign-up
  • No watermark

Hear the real before & after

Clean a recording

Free · up to 3 min

Choose a video or audio file

or drop it here

MP4, MOV, MKV, WebM · M4A, MP3, WAVUp to 2 GB and 60 minutes per file

  • Free full-file cleaning: files up to 5 minutes each, up to 5 a day, one at a time or together.
  • Longer files and batches: preview 60 seconds free, then pay once by card or PayPal. No subscription.
  • Files are uploaded to our server. Uploads are deleted after processing; preview-only uploads within 24 hours. Results expire after 72 hours. Privacy details

    Listen before you upload

    A real desk-fan recording, before and after

    For privacy, the speaker's voice was pitch-shifted before cleaning. The “after” clip is exactly what this tool made from the “before” clip, with the volume matched for a fair comparison.

    Before · fan in the background
    After · processed by this tool
    Before · fan in the background
    After · processed by this tool

    Press play, then switch between Before and After: playback continues from the same moment. The orange band is the fan. Listen during the pauses, then check the words and the tone of the voice. Results depend on your recording.

    Measured

    Measured results on three real recordings

    Run through this site's current processing, volume step included.

    We cleaned three one-take recordings of the same English script by the same speaker, 42 to 46 seconds each, all made near a running fan at different noise levels. The noise floor is estimated as the level of the quietest 10% of 20-millisecond slices of the whole file. Loudness is the integrated loudness (EBU R128) as played on two speakers. Transcription errors is the word error rate of an automatic transcript (OpenAI Whisper) against the script, with numbers written as words.

    Before → after cleaning. Lower is better in both columns.
    RecordingNoise floor (estimate)Transcription errors
    Take 1 · moderate fan noise−40 → −89 dBFS0.0% → 0.0%
    Take 2 · loud fan noise−26 → −65 dBFS5.4% → 2.2%
    Take 3 · loud fan noise−28 → −80 dBFS1.1% → 1.1%
    • The estimated noise floor dropped by 39 to 52 dB. That is not the same as the reduction during speech or a signal-to-noise figure.
    • Loudness stayed within 0.6 dB of the original (−20.3 → −20.9, −14.0 → −14.2 and −22.9 → −22.9 LUFS). When the noise is loud, the model also makes the voice quieter, so the site raises the result back to the original loudness, by up to 12 dB, with a limiter against clipping. On take 2 that meant +9.8 dB, which also lifts what is left of the noise.
    • Transcription errors did not rise; on the noisiest take they fell from 5 to 2 words out of 92. A transcript counts recognisable words, not how natural the voice sounds.
    • The model's raw output is about 0.03 seconds shorter than its input; the site pads it, so your cleaned file keeps exactly the original length.
    • Three recordings are not a benchmark: results depend on your microphone, room and voice. Use the free preview on your own file.

    Noise and loudness measured October 7, 2026 · transcripts October 5, 2026 · model: DeepFilterNet 3

    Guide

    How to remove fan noise without editing the picture

    Four steps, then three things to listen for before you keep the result.

    Updated October 23, 2026

    1. Use the original recording

      Upload an MP4, MOV, MKV or WebM with its audio attached. M4A, MP3 and WAV audio files work too.

    2. Clean one short file free

      For a file up to 5 minutes, choose “Clean it free” to process the entire recording. For a longer file, select a preview start where both speech and the fan are audible.

    3. Compare the voice and the pauses

      Listen to each version at the same playback volume, using the checks below.

    4. Download if it helps

      Free results keep the original video frames and duration. Download within 72 hours. Longer files can be cleaned in full with a one-time payment.

    Three listening checks before you keep the result

    • Between sentences: is the fan less distracting, or does the background pulse as the speaker stops?
    • Quiet words and word endings: are soft sounds still clear, or do syllables sound clipped or watery?
    • During speech: does the voice still sound natural? A quieter pause alone does not make the whole recording better.

    If the noise covered the words in the original, cleaning may not recover them. Keep your original file and compare before replacing it in your project.

    When fan-noise removal is a good fit

    Start with a single speaker who is clearly audible above a steady desk fan, air conditioner or computer hum. The sample above shows one recording, not a guarantee for every room or microphone.

    Gusty wind, traffic, overlapping speakers and room echo need different treatment. Music and singing can be removed along with the noise; this tool is built for speech.

    What changes in the exported file?

    Video frames are copied without re-encoding. Most videos come back as MP4 with AAC audio; WebM files (and MKV files with VP8, VP9 or AV1 video) come back as WebM with Opus audio; ProRes MOV files stay MOV. Audio-only input produces M4A.

    The sound is mixed to mono for cleaning, then copied to both output channels. Only the first video and audio tracks are used; subtitles and extra tracks are not kept. The recording's duration stays the same.

    Reduce fan noise on your next recording

    • Move the microphone closer to your mouth and farther from the fan or computer.
    • If practical, turn off a room fan or AC during the take. Do not block your computer's cooling vents.
    • Record a short test with a sentence and a pause, then listen on headphones before the full take.

    Pricing

    Start free. Pay per job.

    Files up to 5 minutes are free. Longer files and batches cost $2.99 per started hour of audio, paid once, and you can preview 60 seconds of your own file free first.

    Start here

    Free

    Files up to 5 minutes each

    • The whole file, no watermark
    • Up to 5 files a day
    • No account needed
    Choose a file
    Longer files and batches · USD · one-time, no subscription Card or PayPal
    Total audio in one orderPriceBefore you pay
    Up to 1 hour · one file up to 60 min, or up to 30 files$2.9960 s free preview
    Up to 2 hours · up to 30 files$5.9860 s free preview
    Up to 3 hours · up to 30 files$8.9760 s free preview

    $2.99 per started hour of audio in an order. Free: files up to 5 minutes each, 5 per person and network per day; 200 site-wide. Order limit: 30 files, 3 hours and 10 GB in total; 2 GB and 60 minutes per file. No account needed. Full limits and terms · Refund policy

    Files and privacy

    What happens to your files

    Uploads are deleted after processing (or within 24 hours if you only preview). Cleaned results are deleted after 72 hours. We do not use your files to train models. Privacy policy.