Introduction
The Suno stem splitter separates a finished mix into its parts. It works on any mixed song, not only on music made with Suno, and offers three depths: vocals plus background, a full twelve-track split by instrument, or a single part you name together with everything else. Pricing is per song for each mode, so a long track costs the same as a short one.
Key features & highlights
Three depths in one endpoint
two_stems returns a vocal track and a background track, the classic karaoke split. all_stems returns twelve tracks in a fixed order, including backing vocals, drums, bass, guitar, keyboard, strings, synth and fx; parts that are not in the song come back as silent tracks. Sending stem_name instead extracts one named part, from piano or lead vocal to hand claps, and returns the rest of the mix alongside it.
Clean vocals, drums and bass
We built a test mix from known parts, a vocal, a drum kit, a bass line, piano chords and a synthetic string pad, and measured where each part ended up. The vocal, drum and bass stems each contained 99% or more of the right source by spectral energy, with almost no leakage from the others. The synthetic string pad was filed under keyboard rather than strings, a reminder that the split follows how an instrument sounds.
Re-rendered stems, not masks
The stems are rebuilt rather than cut out of the mix, which keeps them free of the warbling that masking leaves behind. The trade-off is that adding the stems back together does not reproduce the original sample for sample, so keep the original mix for any comparison.
Request parameters
| Parameter | Required | Description |
|---|---|---|
audio_url | Yes | The mixed song, MP3 or WAV up to 20 MB. A three-minute WAV is larger than that; send it as MP3. at least 1 characters |
separation_mode | No |
|
stem_name | No | Extract a single part, from |
Pricing
$0.06 / song
Priced per song, not by length: two stems costs least, a single part via stem_name more, and all stems the most.
Use cases
Karaoke and practice tracks
Use two_stems to strip the vocal from a song you have the rights to, then rehearse or record over the background. The vocal stem is useful for checking phrasing or building a lyric video timed to the voice.
Remixes and sample packs
all_stems gives a producer separate drums, bass and harmony to rearrange, re-balance or pass to a mixing engineer. Send stem_name when you only need one part, such as a piano line, and the rest of the mix as a backing track.
Prompt guide & pro tips
Pick the mode for the job
Start with two_stems when you only need the voice out; it is the cheapest mode. Choose all_stems for remixing, and expect several silent tracks when the song uses fewer instruments. Send a specific stem_name instead when a single instrument matters.
Prepare the upload
Send the best-quality mix you have, as MP3 or WAV up to 20 MB; a three-minute WAV is larger than that, so export it as MP3. Avoid recordings of loudspeaker playback, which blur the parts together.
Review before you publish
Solo each stem at normal volume and listen through quiet passages, where leftovers from other parts are easiest to hear. Name the files by stem, and use only songs you have the rights to separate and reuse.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
-H "Authorization: Bearer $SPICY_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "suno/v6/audio-separation",
"input": {
"audio_url": "https://example.com/reference.mp3"
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.


