OpenStudio constructing the production surface
OpenStudio

OpenStudio AI

Generate, separate, and keep producing in the same session.

OpenStudio AI is built around practical production moves: split a mix into workable stems with BS Roformer, turn prompt and lyrics into a WAV through ACE-Step, or use Stable Audio 3 from a separate manual import. The optional setup stays explicit, and the result stays close to the timeline.

Optional AI production tools

  • Stem separation

    BS Roformer 6-stem extraction

    Hand the model any stereo mix and watch it bloom into vocals, drums, bass, guitar, piano, and other. Each lane stays in the project — ready for cuts, replacements, FX chains, or a new arrangement entirely.

    • Optional AI tools setup
    • 6-stem workflow
    • Project-native export
  • Text to music

    ACE-Step prompt and lyrics generation

    Type a prompt, drop in lyrics, choose a seed and duration. The packaged ACE graph runs end-to-end and writes a clean WAV straight onto the timeline — no graph apps, no glue scripts, no detours.

    • ACE-Step 1.5 XL Turbo
    • Prompt + lyrics conditioning
    • Offline after setup
  • Text to audio

    Stable Audio 3 Medium import path

    Stable Audio 3 is a separate optional setup for prompt-based audio, source variation, inpaint, and continuation. It is not downloaded by the BS Roformer or ACE-Step installer path; you import the accepted Hugging Face snapshot deliberately.

    • Manual Hugging Face snapshot import
    • Separate Stable Audio runtime
    • License acknowledgement required

How the local workflow is framed

  • Optional by design

    The base download stays focused. BS Roformer, ACE-Step, and Stable Audio 3 each stay behind explicit setup choices, never bundled by surprise.

  • Offline after setup

    Once the assets are in place, generation validates local files and runs behind offline guards. Your work doesn't depend on a server somewhere having a good day.

  • Quality over quiet fallback

    Full VAE decode is the line. Cross it cleanly, or fail loudly. Nothing is more disrespectful than a silently degraded render pretending to be the real thing.

  • Project material, not demos

    Whatever the AI produces — a separated stem, a generated WAV — lands in the project as a first-class clip. Editable, routable, mixable, finishable.

Practical stem workflows

  • Remix preparation

    Lift the parts you want, leave the rest behind, and rebuild the song around a new center of gravity.

  • Vocal or instrument isolation

    Drop a clean lead into pitch tools, gates, or de-essers without dragging the rest of the mix along for the ride.

  • Arrangement experiments

    Mute lanes, reshuffle sections, and audition alternate structures without rebuilding a project from scratch.

  • Practice and reference work

    Strip a track to its essentials when you need to study a part, transcribe a line, or rehearse against a reduced source.

  • Creative resampling

    Treat each separated lane as new raw material — chop it, layer it, pitch it, run it through anything in the rack.

  • Mix cleanup

    Push a single lane into its own FX chain when the source needs more control than the printed mix can give you.