What is a voice isolator?

A tool that keeps the speech in a recording and removes what is behind it: rain, traffic, hiss, hum, room noise. You upload a file and get an audio file with the voice only.

Is this the ElevenLabs Voice Isolator?

It runs the same ElevenLabs model, Audio Isolation, through fal.ai. Cicadas is an independent site and is not affiliated with ElevenLabs; you need a Cicadas account and credits, not an ElevenLabs subscription or an API key.

How much does the voice isolator cost? Is it free?

It is not free. Isolating a voice costs 25 credits per minute of recording, counted by the second and rounded up to a whole credit: 4 credits for 9 seconds, 13 for 31 seconds, 125 for 5 minutes. The price is shown before you click, and a failed job returns its credits.

Can it isolate the voice from a video?

Yes. Upload an MP4 or WebM up to 80 MB and 10 minutes; the sound track is used and you get an MP3 back. The price follows the length of the sound track, or of the video if that is longer. In our test a 14-second MP4 came back as a clean MP3 in 18 seconds. The video itself is not re-made, so replace the sound in your editor.

How well does it work?

In our tests with rain, hiss, hum and a breaking glass behind an AI voice, the noise in the pauses dropped by 24 to 36 dB, the voice kept its level and a listener that was not shown the scripts wrote down every word. In a fourth test the noise was louder than the voice: the noise still dropped by 22 dB, but a whispered sentence was garbled and several words changed. All four pairs are on this page.

Which files and lengths can I use?

MP3, WAV or M4A up to 20 MB, or MP4 or WebM up to 80 MB, with 5 seconds to 10 minutes of sound. A shorter clip is refused before anything is charged: the model needs about five seconds to work with. Split longer recordings and run the parts one by one. The result is an MP3 (44.1 kHz, mono, 320 kbps) of the same length. In our test a recording just under 10 minutes cost 249 credits and was ready in 98 seconds.

Does it separate vocals from music?

We have tested speech with noise, not singing or songs, so we do not promise stems for music. If you try it on a song, check the result before you use it.

What happens to my recording?

It is stored in your private Cicadas library and sent to the provider for processing; the copy staged for the provider expires within 24 hours. Results can be played and downloaded only by your account. The library holds up to 200 files and 1 GB; each job adds two files, your upload and the result, and files cannot be removed from it yet. Upload only recordings you have the right to use.

Voice Isolator: Keep the Voice, Remove the Noise

Isolate a voice from rain, hiss, hum or street noise with ElevenLabs Voice Isolator. Upload audio or video up to 10 minutes, get a clean MP3. 25 credits a minute.

ElevenLabs Voice Isolator
audio or video
25 credits per minute

Choose a recording or drop it hereMP3, WAV or M4A up to 20 MB · MP4 or WebM up to 80 MB · 5 seconds to 10 minutes

25 credits per minute of recording

ElevenLabs Voice Isolator: 25 credits per minute, counted by the second from the length we measure and rounded up to a whole credit. Charged once when you click; a failed job returns its credits. The result is an MP3 with the voice only.

Your isolated voices

Before and after, made with this tool

Recorded on 2026-10-02. The recordings are our own tests: an AI voice mixed with rain, hiss, hum or a breaking glass. Each “after” is the MP3 exactly as delivered; the last pair shows where it breaks.

Narration over rain
Before
After (voice only)
Ready in 18 seconds. The rain in the pauses dropped by 36 dB; a listener wrote down every word.
MP3 · 8.97s · 4 credits · 01a0fb42-0e50-7c42-bfa3-562064dcb4c3
Instructions over hiss, hum and a breaking glass
Before
After (voice only)
Ready in 25 seconds. The noise in the pauses dropped by 24 dB and the glass is gone; every word was kept.
WAV · 9.9s · 5 credits · 01a0fb42-e5dd-7020-8e7d-7509748c7c12
The sound of a video with rain and hiss
Before
After (voice only)
A 14-second MP4 went in and an MP3 came back in 18 seconds. The noise in the pauses dropped by 34 dB; every word was kept.
MP4 · 14.38s · 6 credits · 01a0fb43-adc7-7a61-ba43-a8600c570044
Where it breaks: noise louder than the voice
Before
After (voice only)
The noise dropped by 22 dB, but words were lost: a whispered sentence came out garbled and “forty years” became “over ten years”. Keep the original when the voice is buried.
MP3 · 30.8s · 13 credits · 01a0fb43-3dfa-78b2-a43f-297d8f4ca8a6

About voice isolator on Cicadas

Upload a recording and get back an MP3 with the voice only. This page runs ElevenLabs Voice Isolator (the Audio Isolation model) on an audio file or on the sound of a video, up to 10 minutes at a time. The price is shown before anything is charged: 25 credits per minute, counted by the second from the length we measure, and a failed job returns its credits. In our tests on 2026-10-02, four recordings of 9 to 31 seconds were ready in 18 to 25 seconds; with rain, hiss or hum behind the voice, the noise in the pauses dropped by 24 to 36 dB and a listener wrote down every word.

Why use Cicadas for voice isolator

Voice isolator for audio and video

Take MP3, WAV or M4A up to 20 MB, or the sound of an MP4 or WebM up to 80 MB. The result is always an MP3 of the same length (44.1 kHz, 320 kbps) with the voice only; a video is not re-made.

A price you see first

We measure the real length of your file, then quote it: 25 credits per minute, by the second, rounded up to a whole credit. A 9-second clip cost 4 credits and a 31-second clip 13 in our tests. You are charged once when you click, and a failed job returns its credits.

Tested, limits included

Rain, hiss, hum and a breaking glass were removed while every word stayed. When the noise was louder than the voice, the result lost words: listen to the last sample pair before you trust it with a badly buried voice.

How to use voice isolator

01
Choose an MP3, WAV or M4A (up to 20 MB), or an MP4 or WebM (up to 80 MB), between 5 seconds and 10 minutes long.
02
Wait a few seconds for the measured length and the price, then click Isolate voice.
03
Listen to the result next to your original and download the MP3.

What you can make

Choose a recording, then finish it in the tool above.

“Clean a voice memo recorded in the rain before you use it as narration.”
“Pull the speech out of a phone video shot on a noisy street.”
“Remove fan hum and hiss from a tutorial voice track before lip sync.”

Voice isolator FAQ

Tested on Cicadas · reviewed 2026-10-02

Recorded on 2026-10-02 through this page's own upload, quote, charge, worker and download path on a local build. Four test recordings we made ourselves (an AI voice mixed with rain, hiss, a 100 Hz hum and a breaking glass): an MP3 of 9 seconds, a WAV of 9.9 seconds, the sound of a 14.4-second MP4 and an MP3 of 30.8 seconds cost 4, 5, 6 and 13 credits and were ready 18 to 25 seconds after the click. Each result was an MP3 (44.1 kHz, mono, 320 kbps) of the same length as its source, downloadable by its owner and refused (401) to a signed-out request. In the pauses of the clean voice, the first three results were 24 to 36 dB quieter than their sources while the speech kept its level within 0.5 dB, and a listener model that was not shown the scripts transcribed them word for word. In the fourth recording the noise was louder than the voice (speech about 5.5 dB below the noise): the pauses still got 22 dB quieter, but a whispered sentence was garbled and several words changed. A recording just under 10 minutes cost 249 credits and was ready in 98 seconds, and a 10-minute 720p video was measured and quoted in 4 seconds. A video is priced on its sound track, or on the video if that is longer: two test videos (2 seconds of picture over 31 seconds of sound, and 30 seconds of picture over 5 seconds of sound) were both quoted at 13 credits, and the provider returned 31 and 5 seconds of voice. The same request sent twice was charged once; a mismatched price, a video without sound and a provider refusal cost nothing. Singing, music stems, recordings over 10 minutes and languages other than English and one Mandarin sentence were not tested.

ElevenLabs Audio Isolation on fal — API specification

Updated Oct 2, 2026 · How we test

Eleven v4 · v4 Turbo · from 1 creditNew

ElevenLabs v4 Text to Speech

Try Eleven v4 and v4 Turbo online: paste a script, pick one of 20 voices and get an MP3 in seconds. Audio tags like [whispering] work. From 1 credit.

Tested 2026-10-02

Sync Lipsync 2 · 12.5 credits per secondNew

AI Lip Sync

Make a talking-head video match new speech and get a private MP4 back. This AI lip sync tool runs Sync Lipsync 2 and costs 12.5 credits per second of video.

Tested 2026-09-29

Sound effects for your clips

ElevenLabs Sound Effects v2 · from 1 creditNew

AI Sound Effect Generator

Describe a sound and get an MP3 in seconds: footsteps, whooshes, ambience or loops up to 22 seconds, made with ElevenLabs Sound Effects v2. From 1 credit.

Tested 2026-10-01