Skip to content

Remove background noise.Keep your voice.

Fans, traffic, a busy room, old tape hiss. Drop an audio or video file and Shhh takes the noise out from behind the speech. The strength slider decides how far it goes.

  • 1 free hour every month
  • Audio or video
  • Trained on 102 languages

Real recordings

Hear the noise go. Decide how far.

Real recordings, with the noise that was really there. Hold the button on the video to hear the Shhh version, let go for the original, or drag the strength while it plays.

Busy outdoor fair

0:00 / 0:00

As recorded

The strip is the sound: the haze is noise, the bright lines are the voice. Click it to jump. noise voice

Festival interview in LimaBusy outdoor faires

0–50%

Natural

Your own voice, with the noise turned down.

50–100%

Studio

The voice is rebuilt for a studio sound.

Source: Rosario MQ, CC BY-SA 4.0 (this processed clip is shared under the same license). Trimmed; audio cleaned with Shhh.

How it works

How to remove background noisefrom audio

  1. Drop your file

    Audio or video, or record from your mic. Try a file up to 3 minutes long without signing in.

  2. Pick a strength

    Natural turns the noise down around your own voice. Studio rebuilds the voice for a studio sound.

  3. Listen, then download

    Flip between before and after. WAV or MP3, same sample rate and channels as your file.

What it handlesand where it struggles

Works well on

  • Steady noise: fans, air conditioning, computer hum, room tone
  • Street and traffic noise
  • Crowds, fairs and busy rooms, like the festival clip in the demo
  • Hiss and crackle on old recordings
  • A voice far from the mic, with room noise around it

Harder cases

  • People talking in the background. They are voices too, so some of what they say may stay.
  • Noise so loud you cannot make out the words yourself. Speech that is not in the recording cannot be brought back.
  • Very hard recordings at high strength. The voice is rebuilt, and a sound can come out different. Listen to the preview, and lower the strength if a word sounds wrong.
  • Music under the voice goes with the noise. There is no separate music control yet.

Tips

  1. Start with the 30 second preview. It plays the same moment at two strengths, so you can choose by ear before you use any of your minutes.

  2. If the recording is already fairly clean, a lower strength is often enough.

  3. Making a podcast? Turn on Even out loudness (-16 LUFS) so it plays at the level podcast apps expect.

  4. Before the next recording, run Mic Check. It measures your level, clipping, room noise and echo in a few seconds.

Questions

Something not answered here? Write to us.

support@shhh.sh
Is it free?

Yes, with a free account: whole files up to 30 minutes, 1 hour of audio every month, no card needed. Without an account you can try the first 30 seconds of a file up to 3 minutes long. You can pay once for a longer file, no account needed, and Pro covers 30 hours a month. The pricing page shows the prices in your currency.

Will my voice sound robotic?

That is what the strength slider is for. Up to 50%, Shhh does not rebuild your voice at all: it uses its own result as a guide and turns down everything in the original recording that is not your voice. Above 50%, it moves toward a rebuilt studio voice. If the result sounds processed, pull the slider back.

Can it change the length or add sounds?

The length, no. The cleaned file is exactly as long as the original, sample for sample, so it lines up with your video or your edit. At high strength the voice is rebuilt, and on very hard recordings a sound can come out different, so listen to the preview before you clean the whole file.

Can I clean the sound of a video?

Yes. Drop an MP4 or MOV and you get the same video back with clean audio. The picture is copied as is, never re-encoded.

Does it work in my language?

Most likely. The model behind Shhh learned from speech in 102 languages, and we test English, Japanese and Spanish ourselves before each release.

What happens to my file?

Files are deleted after 7 days and are never used to train models. Your list of past files is kept in your own browser.

Drop the noisy take. Keep the voice.

Try it free

Try 30 seconds of a file up to 3 minutes long, no sign in.