Sample packs: isolate before you demo

Isolate. Do not clone the demo voice.

A sample pack is stems, one-shots, and a demo mix that proves they sit together. The demo vocal is evidence, not a product you should duplicate. Pull the voice out of the mix when you need a dry vocal or a cleaner listen. Do not train a reusable clone off the singer you hired for the walkthrough.

Isolation is a mix job

Isolate vocals from a track runs isolate_audio on any mixed audio_url. It strips background music and instrumentation and returns a vocal track. It is not limited to files you generated in Versely. Quality follows the source mix: a clean demo separates; a crushed preview will not.

That is the tool for "I have a mixed pack demo and I need the vocal dry for a second video, a translation prep, or a cleaner store listing." Billed per clip. Ask the agent for estimate_cost on a long file. Reattach with attach_audio_to_video if the next object is a picture.

Split a song into stems is a different object. separate_music_vocals only works on a track you already generated in-app with generate_music or extend_music, referenced by taskId and audioId. It is the karaoke/remix split for a Suno-grade generate, not a general unmix of someone else's WAV.

If the pack demo was mixed outside Versely, isolate. If the demo was a Versely generate, you may split stems from that generate. Do not confuse the two buttons.

Cloning is a new voice

Clone my voice from a recording is clone_voice_from_audio: attach a sample, name it, get a reusable voice_id. From then on the agent can read new scripts in that identity. The voice tool's own rule is you clone voices you own or have explicit consent to use.

A pack demo vocal is usually a hired or featured performance for that mix. Cloning it gives you a second product — a TTS identity — that the pack page was not selling. It is also how you accidentally put a singer into every future VO. Isolation does not create that object. Cloning does.

Do not "make the demo speak a new verse" by cloning. Write a new vocal on a voice you have rights to, or keep the original performance and isolate it when the listing needs a cappella.

Demo the pack you actually sell

AI music generator is for original tracks from a prompt: Suno-grade generation, full songs or instrumentals, commercial use on a paid plan. That is how you score a new demo in a genre, not how you unmix a pack you already mixed.

The listing workflow that stays honest:

1. Mix the pack. Export the demo.
2. If the store needs a dry vocal or a bed without the singer, isolate (or split, if the demo was an in-app generate).
3. Do not clone the demo singer into a reusable voice to "localise the walkthrough."
4. If you need a new spoken walkthrough, generate speech on a voice you own.

The pack is the samples. The demo is proof. Isolation is a file transform. A clone is a new instrument with a name. Selling the first while quietly shipping the third is how a sample shop becomes a voice shop without meaning to.

FAQ

Can I isolate any uploaded demo, or only Versely tracks?

Any audio_url. Isolation is general-purpose. Stem split (separate_music_vocals) is the one that only works on a prior generate_music / extend_music result.

Why not clone so I can redo the demo in Spanish?

Because cloning is a new voice identity, and the demo singer is not that identity unless you have the right to clone them. Isolate the original, or record / generate a walkthrough on a voice you own. Translation of a video is a different editing job.

Does isolate give me the instrumental too?

isolate_audio returns the vocal with music removed. For vocal and instrumental stems of an in-app generate, use split-a-song-into-stems. For an outside mix, isolation is the vocal tool — do not expect a full stem session from it.

Should I generate a new Suno demo instead of isolating?

If you need a new original track, yes — that is the music generator. If you need to show the pack you already mixed, isolate the mix you have. Generating a cousin track is a different SKU.