Audio · SeparationAvailable

Vocal Separation 2.5

zilapi/vms-2.5 · by InverseAI

Splits a song or mix into a vocal track and an instrumental track. Both are delivered as MP3, up to 320 kbps.

Vocal separation API

POST /v1/audio/separate takes a song or mix and returns two files: the vocals on their own and the instrumental without them. Both stems are included in one price, so you can build karaoke tracks, remixes, a cappellas or a vocal remover into your product without running a GPU yourself.

What developers use it for

Make karaoke and backing tracks by removing the vocals
Extract a cappella vocals for remixes, covers and samples
Remove the vocals from songs used as background music
Feed clean vocals into lyric transcription or pitch analysis

Create a job

One request: POST /v1/audio/separate with the audio's input_url (or the file itself as multipart, up to 50 MB) and a webhook_url. We fetch it, run zilapi/vms-2.5 and POST job.succeeded to your URL. The quickstart has every step.

curl -s https://api.zilapi.com/v1/audio/separate \
  -H "Authorization: Bearer $ZILAPI_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{"input_url": "https://example.com/demo-song.mp3",
       "model": "zilapi/vms-2.5",
       "webhook_url": "https://your-app.example.com/hooks/zilapi"}' \
  > job.json
JOB_ID=$(jq -r .id job.json)

Outputs

A succeeded job has 2 files in output.files. Download each with GET /v1/jobs/{id}/output?name=…. <stem> is your input's name without its extension.

Outputs of zilapi/vms-2.5
name	Filename	Contains
vocals	<stem>-vocals.mp3	The voice(s) only, MP3, up to 320 kbps
instrumental	<stem>-instrumental.mp3	Everything except the voice(s), MP3, up to 320 kbps

for NAME in vocals instrumental; do
  curl -s "https://api.zilapi.com/v1/jobs/$JOB_ID/output?name=$NAME" -H "Authorization: Bearer $ZILAPI_KEY" > output.json
  curl -o "$(jq -r .filename output.json)" "$(jq -r .url output.json)"   # <stem>-$NAME.mp3
done

Limits and formats

Provider	InverseAI
Input formats	Common audio: WAV, MP3, M4A/AAC, FLAC, OGG and more, or the audio of a video
Max duration	6 h per file (21,600 s)
Max file size	5 GB (5,368,709,120 bytes)
Direct upload	Up to 50 MB in the job request itself (multipart)
Output	2 files, MP3, up to 320 kbps
Billing	Per second of input, 1-second minimum per job, both outputs included
Price	$0.033 / audio min = $0.00055 / s
Retention	Input deleted 3 days after it is ready; outputs 7 days after the job succeeds

Parameters

Field	Type	Description
input_url	string	An https URL we fetch (up to 5 GB, 6 h). Send this or `file`.
file	string	A ready `file_…` id from `POST /v1/files`; or, in a multipart request, the audio itself (up to 50 MB).
model	string	Optional. `zilapi/vms-2.5`, the default for Vocal Separation.
webhook_url	string	Optional. An https URL that receives this job's `job.*` event, signed. See Webhooks.

Vocal separation API: questions

How much does the vocal separation API cost?

$0.033 per audio minute of input, billed per second. Both outputs, vocals and instrumental, are included.

What do I get back?

Two MP3 files, up to 320 kbps: vocals and instrumental. Download each with GET /v1/jobs/{id}/output?name=vocals or name=instrumental.

How long can a song be?

Up to 6 hours or 5 GB by URL, or 50 MB uploaded directly in the request.

Compare every model on the pricing page, read the step-by-step guide, or start with the quickstart.