Voice isolator

curl https://api.iodine.dev/v1/isolate \
  -H "Authorization: Bearer $IODINE_API_KEY" \
  -F file=@interview.mp3 \
  -o interview-voice.wav

The response is the audio itself: just the voice, as mono 44.1 kHz WAV, with music and background noise removed.

Use it to clean up a recording before cloning a voice from it, to tidy podcasts and voice-overs, or to take the music out from under someone speaking. To try it without writing code, open the console's Voice isolator: upload or record audio, hear the original and the result side by side, and download it, at the same price.

#Request

Multipart form fields:

The response's Content-Type is audio/wav or audio/mpeg. Its headers also include:

/v1/audio-isolation is the same endpoint at ElevenLabs' path, and it accepts their audio field name.

#Examples

import os, requests

with open("interview.mp3", "rb") as f:
    r = requests.post(
        "https://api.iodine.dev/v1/isolate",
        headers={"Authorization": f"Bearer {os.environ['IODINE_API_KEY']}"},
        files={"file": f},
        data={"format": "mp3"},
    )
r.raise_for_status()
open("interview-voice.mp3", "wb").write(r.content)

const form = new FormData();
form.append("file", new Blob([await fs.promises.readFile("interview.mp3")]), "interview.mp3");
const res = await fetch("https://api.iodine.dev/v1/isolate", {
  method: "POST",
  headers: { Authorization: `Bearer ${process.env.IODINE_API_KEY}` },
  body: form,
});
await fs.promises.writeFile("interview-voice.wav", Buffer.from(await res.arrayBuffer()));

#What it removes, and what it doesn't

• Music under speech: removed, even when it's loud.
• Background noise: fans, traffic, hum, hiss.
• Echo isn't removed: the voice comes back sounding like the room it was recorded in. Record somewhere soft-furnished if that matters.

If several people talk at once, you get all of their voices back: it keeps voices and removes everything else, but it doesn't pick one speaker. Singing counts as voice, so it's kept.

#Pricing

$0.001 per minute of audio ($0.06 an hour), measured from the recording you send, to the millionth of a dollar. Requests that fail on our side cost nothing. It draws on the same credit balance as text to speech, and shows in the console's usage as isolate.

#Your audio

Recordings are processed in memory on our own servers and never stored, whatever your Zero Data Retention setting; we keep only the request's metadata (length, cost). See Billing and data.