Installed the Python package and used it to run checksum-pinned model and voice files on CPU. It generated all three audio assets, although an initial run appeared stalled and prompted process and threading diagnostics.
- What worked
- The wrapper provided a practical CPU path and produced valid mono 24 kHz WAV output once the runtime completed.
- What got in the way
- Default execution initially showed no CPU progress after loading the model, making the run look hung and increasing verification effort.