Skip to main content

elevenlabs/audio-isolation/audio-to-audio

Use ElevenLabs Audio Isolation to process speech in an audio or video source. The request takes audio_url and returns a cleaned MP3, with billing based on source seconds.

From $0.0022/s

Workbench

Input

Audio to audio
affects price1/1
1 / 1 files
Required URL of the source audio or video, within 500 MB and one hour. Upload MP3, WAV, MP4 or WebM on the page; use the API URL field for other listed formats. The source duration sets the billed seconds. at least 1 characters
Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run

Related models and next tasks

Missing the model you need?

ELEVENLABS AUDIO ISOLATION

Bring speech forward from a noisy recording.

Key numbers

Maximum duration
1 hour
ElevenLabs lists a one-hour limit for each source recording.
Maximum file size
500 MB
ElevenLabs lists a 500 MB limit for each submitted file.
Returned format
MP3
In our tests, audio and video inputs returned mono MP3 audio.
Noise floor in our test
−35 dB
In our tests, a noisy voice message went from about −24 dB to about −35 dB.
File retention
14 days
Generated files are deleted after the 14-day storage period.

Before you spend

All modes

Generated files remain available for 14 days and are then deleted. Save any results you need before that period ends.

Audio to audio

The source must fit within one hour and 500 MB. Every result is MP3 audio, including video submissions. Cleanup strength, loudness and output format cannot be adjusted. ElevenLabs says the capability is not specifically optimized for vocal isolation from music.

What it does

Measured speech cleanup

See the noise floor change

Before and after spectrograms of the noisy voice message processed with ElevenLabs Audio Isolation.

In our tests, the noise floor of a noisy voice message fell from about −24 dB to about −35 dB. The spectrogram pairs that recording with its cleaned result for comparison.

Pauses go quiet

Gaps between words fall away

Two waveforms of the same voice message, the noisy source above and the isolated speech below, where the pauses between words have gone quiet.

In our tests, the pauses in the noisy voice message dropped close to silence while each spoken phrase stayed where it was. Compare the two waveforms to see where the hiss and hum used to sit.

Video accepted

Take the cleaned sound into your edit

A video submission returns MP3 audio without a picture. In our tests, a 6.04-second video produced a 6.04-second audio track.

Where it fits

  • Voice notes from busy places

    Try cleaning a spoken message recorded near traffic or a crowded platform, then compare the words and pauses with the original.

  • Interview excerpts

    Process a spoken interview recorded with unwanted ambient sound and review the cleaned MP3 before including it in a podcast.

  • Speech in a video edit

    Submit a video with noisy dialogue, then bring the returned audio into your editor alongside the source picture.

Model reference

ElevenLabs Audio Isolation API reference

ElevenLabs Audio Isolation API specifications

API name
elevenlabs/audio-isolation/audio-to-audio
Source files
Audio or video; maximum 1 hour and 500 MB
Result audio
MP3; mono, 44.1 kHz, about 320 kbps in our tests
Request setting
audio_url; no strength, loudness or format controls
Duration billing
Source seconds rounded up; minimum 1 second, maximum 1 hour

How to call the ElevenLabs Audio Isolation API

  1. Choose a source file

    Use the page upload for MP3, WAV, MP4 or WebM. For another supported format, submit its URL through the API.

  2. Submit its URL

    Set audio_url to the recording you want to process. Check that the file is within both limits: 500 MB and one hour.

  3. Listen to the MP3 result

    Compare the cleaned speech with your source before using it in an edit. Download generated files within 14 days, before they are deleted.

ElevenLabs Audio Isolation API questions