Browse this guide

Neural Analog guide

Core task

Split a song into stems

Extract vocals, drums, bass, and instruments, or describe one sound for SAM Audio. Completed stem rows appear below the full song.

Extract standard stems

  1. Open Stems and choose 4 stems for vocals, drums, bass, and other, or select another preset.
  2. The imported file is already selected under Source audio. Change it only to split another version.
  3. Leave Process stems off for separation alone.
  4. Review the estimate and choose Start Extraction.
Studio Stems panel with preset, source, Process stems, Import stems, and Start Extraction controls
Choose a preset and source, decide whether to process every output stem, then start extraction.

Extract one described sound

  1. Choose the Custom/SAM Audio option and enter a short prompt such as “lead vocal” or “electric guitar solo.”
  2. If the sound occurs only in part of the song, hold Ctrl/Command and drag over that interval to add a prompt region.
  3. Test the prompt with Small or Base before using the higher-cost Large model.

This demo shows how SAM Audio isolates a sound described with a text prompt.

Understand Process stems

When enabled, Process stems runs the selected enhancement, mastering, or transcription model after separation and creates a processed version for every non-silent output stem.

Import stems you already have

  1. Open Stems and choose Import stems.
  2. Select one or more stem files. Each file becomes a stem row; no separation model runs.
  3. The files must already share the same start time. Studio does not time-align unrelated bounces.

Review and export

  1. Switch playback to stems and Solo each row to check the separation.
  2. Clear Solo and play all stems together to check the reconstruction.
  3. In Export, choose Combined Stems for the current mix or Individual stems for separate files.

Next steps