Noise remover

What the noise remover changes

Speech and music need different treatment

Start by choosing the recording you actually have. A spoken interview with fan noise needs a different treatment from a stereo song with tape hiss. The noise remover checks the file and provides a speech route and a music route rather than passing every sound through a voice model. Speech is cleaned on a server after the run is confirmed. Music uses a noise profile and processing in your browser. This distinction matters because a speech model can mistake cymbals, sustained notes and reverb for unwanted sound. Use this workflow to lower an identifiable floor, then listen for anything useful that followed it out.

Hear what comes out

A cleaner file can seem better simply because its level changed. The noise remover compares Original and Cleaned at matched loudness, so a louder voice does not hide a damaged consonant or watery tail. Listen through the end of a sentence and the following pause, then try a louder passage. The removed-sound lane gives another check: it should be dominated by the unwanted background rather than whole syllables or an instrument. On the speech result, you can mix some original sound back underneath. Lower the reduction when a perfectly quiet pause makes the recording sound less believable.

Protect quiet detail

Steady noise is easier to describe than a second person speaking over the first. The noise remover can lower a fan, hiss or room bed, but it does not identify the one person you intended to keep. Nearby talkers may remain, and loud noise that masked a word cannot be turned back into a known original word. Music mode is most useful when the file contains a short noise-only passage. That supplies a reference for what to reduce. When noise and music occupy the same frequencies throughout the track, use a gentler amount and accept some remaining hiss.

Where a quieter floor helps

• 

  Spoken interviews

  An interview often moves between clear answers, quiet thinking and noisy interruptions. Use the noise remover on a representative excerpt before committing a long session. Check both the closer and the more distant speaker; a setting that flatters the presenter can make the guest sound thin. Pay attention to the start of words after a pause, where aggressive reduction can feel like a gate opening late. This page handles background cleanup, while the podcast audio enhancer is a better next step when you also need click repair, breath control and delivery loudness. Keep the unprocessed recording so another edit remains possible.

• 

  Archive transfers

  A cassette or vinyl transfer can have useful stereo ambience as well as unwanted hiss. In Music mode, the noise remover stays on your device and learns from the quiet parts of the file. If there is a clean lead-in or run-out, use it as the reference instead of a section containing a quiet sustained instrument. Listen to cymbal fades, string decays and the space around the recording. Those sounds can resemble the noise floor. The cleaned copy should make an archive easier to hear without pretending to restore frequencies that were lost before digitization. Save a separate cleaned copy and preserve the transfer.

• 

  Voiceover preparation

  For narration, judge cleanup at the level of the final listening experience. A faint room bed that disappears under a voice may not need maximum suppression. The noise remover is useful before an editor assembles takes, because a consistent background makes joins easier to hear and compare. It does not remove mouth clicks or decide where breaths belong in a performance. For those jobs, use the dedicated tools and review their marks. After the noise remover, compare a quiet phrase with a forceful one through headphones. If the voice changes character between them, restore some original sound rather than pushing for a silent waveform.

Choose the route, then compare

1. 1

   Load and confirm the recording

   Drop a supported audio file and read the file information before starting. Choose Speech for spoken material and Music for a song or instrumental recording; if the automatic reading is unclear, answer according to the content you want to preserve. The noise remover does not upload a speech file merely because you selected it. Review the confirmation and the available run count first. For music, prepare to audition the local result without a cloud run. Noise reduction cannot improve the source format itself, so use an original WAV or FLAC when available rather than repeatedly saving a compressed copy between cleanup attempts.

2. 2

   Compare difficult moments

   After processing, switch between Original and Cleaned while playback continues through a noisy pause and into speech. Check a soft ending such as an s or f, then a sudden consonant and a breath. The noise remover should leave the words easy to understand, even when some background remains. In Music mode, also check an exposed bass note and the last cymbal of a section. Use a noise-only reference when one exists, and avoid teaching the profile from music. Treat this as an auditioning step: a smaller reduction that survives close listening is often more useful than the quietest available result.

3. 3

   Save a checked result

   Download the version you actually compared, then listen to a short section of that saved file in your normal player. Keep its name distinct from the original so the untreated source remains easy to find. The noise remover is one stage in an editing workflow, not a substitute for checking joins, final level or permissions to publish the recording. If the sound needs room treatment rather than a lower background floor, compare the echo or reverb tool before adding more noise reduction. Return to the noise remover only when the remaining problem is noise; repeatedly processing an already cleaned result can accumulate dullness and artificial texture.

Common questions

Learn how

Guides and tutorials that use this tool, with real screens and real numbers.

• GuideBackground noise removerHow to clean up a noisy voice recordingHum, fans, traffic, hiss and room echo each need a different fix. How to tell them apart, which tool takes out which, and what to expect from the result.