Choose the right stems for singing, practice and remixing

A singer preparing an accompaniment and a bassist studying a bass line need different outputs. For the singer, 2-track separation provides vocals and accompaniment. For the bassist, 4 tracks provides vocals, drums, bass and other. Start by naming the part you want to hear, practice or edit; that makes the mode choice straightforward.

AI separation estimates sources from a finished mix, giving you individual parts to audition and use in your workflow. Sounds that overlap in the mix can remain partly shared between those outputs. Listen both for the part you want to keep and for sounds from the rest of the arrangement. Original studio stems come from the recording session; separated tracks are useful estimates made from the finished recording.

The 6-track mode adds separate guitar and piano groups to vocals, bass, drums and other. Choose it when those instruments matter to your practice or edit. The 12-track mode also distinguishes synth, strings, percussion, brass, woodwinds and effects; it is available only for supported songs generated inside Tunenoodle, not arbitrary uploaded recordings.

Separation requires signing in, a ready source in your library, sufficient credits and an available processing service. The source must be no longer than 8 minutes. Check whether your source actually contains the instruments you want to isolate. A group can be quiet when its instrument is absent or masked. “Other” collects the sounds outside the named groups and is worth auditioning when you are looking for a texture or secondary part.

The current rates are 4 credits for 2 tracks, 10 for 4 tracks, 10 for 6 tracks and 12 for 12 tracks, per billable minute. Specialized separation modes use 15. Billable minutes are whole source minutes rounded down, with a minimum of 1: max(1, floor(seconds ÷ 60)). Use the completed whole minutes when calculating your estimate.

A 3:45 source counts as 3 minutes: 12 credits for 2 tracks, 30 for 4 or 6, and 36 for eligible 12-track separation. A 0:45 source counts as 1 minute, so 2 tracks costs 4 credits. A 4:00 source counts as 4 minutes. The button quote is the final check before a new submission; these examples describe the implementation reviewed on September 25, 2026.

1. Keep your original file and choose the clearest available source. Process that source directly to avoid additional lossy conversions.
2. Sign in, select the ready song or upload a source, and choose 2 tracks for vocals plus accompaniment. Read the displayed credit quote before submitting.
3. When the result is ready, listen to the verse, densest chorus and final sustained note. Check the accompaniment for recognizable lyric fragments and the vocal track for drum or synth spill.
4. Download the required result and test it at the actual practice volume. If the backing track supports your singing, a little residue may be acceptable; an exposed vocal remix may need substantially cleaner material.

Compare the same passage with the original, especially where reverb, doubled vocals or similar-sounding instruments overlap. For a vocal, listen for clear consonants and a complete final note. For accompaniment, check that the rhythm stays steady and that any remaining lyric fragments sit comfortably behind your own voice. This gives you a concrete standard for rehearsal, study or editing.

If a group combines parts you want separately, choose a mode that names those instruments and review its credit quote before submitting again. Work from the clearest original available. Keep playback levels similar when comparing, since turning up an isolated track also makes residue more noticeable. For an exposed solo edit, choose the cleanest passage; for practice, a part that makes the musical line easy to follow may already serve the task well.

Which mode should I start with? Choose 2 tracks for an accompaniment, 4 for the core rhythm groups, and 6 when guitar or piano needs its own file. Use 12 tracks when an eligible generated source and the additional instrument groups fit your edit. Track count describes the output groups; assess sound quality by listening to your source.

How do I work with room sound in a vocal? Expect some of the room or effects in the original mix to remain. Audition the vocal in the intended arrangement, where those tails may help it blend. A dry, exposed production needs closer attention to each phrase.

How should I label the files? Keep the song, mode and part in the filename, and describe the outputs as separated tracks when sharing a technical comparison. Save the original alongside them so you can return to the source for another edit.