Audio Separation

Vocal Separator API — isolate vocals or instrumentals

Two specialized models — speech and music — that pull clean vocal tracks or backing tracks out of any mix, with minimal artifacts.

AudioSeparation

Before & after

See the difference

One mixed file goes in — two clean stems come out. The waveform shows the split in action.

INPUTOne mixed audio file
Mixed input
Voice + instruments in a single track
OUTPUTTwo clean stems
VocalsVOCALS
Lead voice, reverb and bleed removed
InstrumentalOTHER
Music bed, phase-cancelled

What it does

Vocal Separator pairs two single-purpose models. Speech mode is built for dialogue, podcasts and voiceover: it keeps intelligibility while removing music beds and effects. Music mode separates lead vocals from instrumentals for karaoke, remixes and stems. Both run without phase cancellation tricks, so the isolated track stays clean enough for broadcast.

Full API reference
Python
import requests

url = "https://api.mlslabs.io/v1/vocal-separator/jobs"
resp = requests.post(
    url,
    headers={"X-API-Key": "YOUR_API_KEY"},
    json={
        "input": "s3://bucket/episode-05.wav",
        "output": "s3://bucket/stems/",
        "mode": "speech",        # or "music"
        "tracks": ["vocals", "instrumental"],
    },
)
job = resp.json()
print(job["job_id"])

Key features

Why teams pick this API

Speech mode

Built for dialogue, podcasts and voiceover — keeps intelligibility while removing music beds and effects.

Music mode

Separates lead vocals from instrumentals for karaoke, remixes and stems.

Artifact control

No phase cancellation, no musical noise under the vocal — clean enough for broadcast.

Streaming-friendly

Low-latency segment processing for live and interactive audio workflows.

Use cases

Built for real workflows

Karaoke & backing tracks

Produce instrumentals and isolated vocals for consumer products in one request.

Podcast & interview cleanup

Remove music beds and room tone while keeping dialogue fully intelligible.

Dubbing & voiceover prep

Isolate the original voice for reference or replacement during localization.

Remix stems

Grab the vocal of one track and the instrumental of another for mashups.

Pricing

Credit-based pricing

Every request draws from your credit pool. This API bills 12 credits per minute of audio.

Billing unitCost in creditsNotes
Processing time12 creditsPer minute of audio
Failed jobs0 creditsCredits auto-refunded
Batch processingSame rateOne queue, per-file callbacks

See credit packs & full pricing

Learn

Guides, tutorials & comparisons

Browse all tutorials

FAQ

Common questions

What is the difference between Speech and Music mode?

Speech mode is optimized for dialogue-like content; Music mode for songs. Picking the right mode meaningfully improves separation quality.

Can it run in real time?

Yes, in streaming mode — designed for live and interactive audio.

What about full stem separation?

For multi-stem output (drums, bass, piano, guitars, environmental sound), use the Source Separation API instead.

Does it work on mono files?

Yes, mono and stereo inputs are both supported.