Workflow

Revoice a Song

Re-sing an existing song in a different voice. Soundverse transcribes the original lyrics, separates the vocal from the instrumental, regenerates the vocal in the voice you describe while keeping the lyrics and melody, and layers it over the original instrumental.

Quickstart
Request fields
How it works
Results and playback
Pricing

Dedicated route only

This workflow uses POST /v1/revoicings. It runs a fixed multi-step pipeline; there is no generic tool_id for it.

Quickstart

POST/v1/revoicingsrevoice a song

curl -X POST https://apiv2.soundverse.ai/v1/revoicings \
  -H "Authorization: Bearer sksoundverse_..." \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: revoicing-001" \
  -d '{
  "song_file_id": "018f0000-0000-7000-8000-000000000002",
  "voice_style": "male baritone, warm and breathy",
  "license": "royalty_free"
}'

The song_file_id must be an active uploaded audio file owned by the authenticated enterprise account.

Getting a song_file_id

Upload the file first via POST /v1/files — either {"source_url": "..."} if you already host it, or a multipart/form-data file part to upload bytes directly. Both return a file_id owned by your account, ready to pass here.

Request fields

Field	Type	Required	Default	Description
song_file_id	uuid	Yes	—	ID of the song whose singer should change, an audio file owned by the authenticated enterprise account (upload it first via POST /v1/files). Maximum 60 MiB.
voice_style	string	Yes	—	The singer to regenerate the vocal as: voice/gender, timbre, and delivery (e.g. 'male baritone, warm and breathy'). The original lyrics are transcribed from the track and the original melody and phrasing are preserved; only the voice changes.
license	license tier	No	"royalty_free"	License tier to price and attach to every step of the pipeline.

How it works

Each request runs 6 steps in order:

Transcribing the original song’s lyrics.
Separating the song into vocal and instrumental stems.
Regenerating the vocal in the requested voice, keeping the original lyrics and melody.
Making sure the new vocal doesn’t run longer than the original track.
Layering the new vocal over the original instrumental.
Rendering the revoiced song into one playable file.

Progress streams over GET /v1/generations/{task_id}/stream as each step starts and finishes.

Notes

Use a song with vocals: its lyrics are transcribed and the vocal is regenerated in the voice you describe, trimmed to the original’s length, and layered over the original instrumental.

Results and playback

Poll GET /v1/generations/{task_id} until the task is completed. The finished result is added to the authenticated account’s Soundverse Library. Enterprise API outputs are hidden from public/profile surfaces by default.

For API-only download, resolve the output’s file_id with GET /v1/files/{file_id}/url and use the returned short-lived signed URL.

Pricing

This workflow is priced pass-through: you are billed for each step the pipeline runs, at that step’s enterprise rate, in the license tier you requested. The pipeline itself adds no separate fee, and your billing ledger shows one entry per step. The steps that carry the cost are:

Lyric transcription
Stem separation
Vocal regeneration (similar singing)
Duration trim
Mixing
Final render

There is no fixed per-request price; the total depends on the input and on which optional steps run.

← Drumless Practice MixAdd an Instrument Layer →