CREATOR STUDIO - LOCAL AI INSTRUMENTAL PREPARATION (prototype)

Requires Windows, Python 3.11 with the py launcher, and several GB free disk/RAM.
This runs Demucs on your computer. Nothing is uploaded. CPU separation may take
longer than the track. It is preparation, not real-time processing of Spotify.
Use a local WAV or FLAC file (at most 30 minutes and 512 MB). Original audio is never overwritten.

1. Extract this ZIP into a writable folder. Review the PowerShell script.
2. In Windows PowerShell, open that folder and run:
   powershell -NoProfile -File .\Prepare-Instrumental.ps1 -Install
   Downloads CPU PyTorch and Demucs into %LOCALAPPDATA%\CreatorStudio\VocalTools.
   If execution is blocked, follow your local script policy. The script does not
   change execution policy or require administrator rights.
3. Run:
   powershell -NoProfile -File .\Prepare-Instrumental.ps1 -Track "C:\Music\song.wav"
   The first separation also downloads the htdemucs model. Keep that model cache
   for offline use. Later runs with the same cached model need no connection.
4. In DJ Studio choose + Add local files in the Music library and select the
   generated no_vocals.wav. Load or queue it on either deck like any other track.

Prepared stems may contain vocal remnants or artifacts. The slider labeled
Live vocal reduction is a separate center-reduction effect, NOT Demucs. It can
remove instruments too and performs poorly on mono or heavily reverberant vocals.
Clips saved from the instrumental use that instrumental audio. Live effects are
not baked into source clips. Master set recording captures what you hear.

Pinned packages: Demucs 4.0.1, PyTorch/torchaudio 2.5.1 CPU, soundfile 0.13.1,
numpy <2. Third-party software/model licenses apply; see their upstream projects.
Native Windows installation/separation needs validation on the user's Windows PC.
