Python Runtime Setup

The Auto-Chart feature requires Python. Editing charts does not — if you only edit, you can ignore this page.

Packaged builds (default)

OCTAVE bundles a Python 3.11 runtime. The first time you click Auto-Chart, OCTAVE downloads:

• A standalone Python 3.11 build (~25 MB)
• The required ML model weights (~250 MB total: Demucs + basic-pitch + whisper)

These are cached locally and never re-downloaded:

Platform	Cache path
Windows	%APPDATA%\octave\python-runtime\
macOS	~/Library/Application Support/octave/python-runtime\
Linux	~/.config/octave/python-runtime\

To wipe the runtime (e.g. corrupted download): delete that folder. OCTAVE re-downloads on next run.

Dev builds (running from source)

When you run npm run dev, OCTAVE looks for a Python interpreter in this order:

1. OCTAVE_STRUM_PYTHON environment variable (absolute path to a python.exe)
2. Project-local .venv\Scripts\python.exe (Windows) / .venv/bin/python (Unix)
3. Bundled runtime
4. System python3.11

The recommended dev setup is a project-local virtualenv:

# from repo root
python3.11 -m venv .venv
.venv\Scripts\activate          # Windows
# source .venv/bin/activate     # macOS / Linux
pip install -r resources/strum/requirements.txt

Then just run npm run dev — OCTAVE picks up .venv automatically.

Custom Python interpreter

In Settings → Auto-Chart → Custom Python interpreter, point to any Python 3.11 binary that has resources/strum/requirements.txt installed. This sets OCTAVE_STRUM_PYTHON for you.

GPU acceleration

The bundled runtime detects:

• CUDA on Windows / Linux (NVIDIA only) — set automatically if nvidia-smi is on PATH.
• Metal on macOS — used automatically.
• CPU fallback — always available; slower but works on every machine.

To force CPU even when a GPU is available, tick Settings → Auto-Chart → Force CPU.