turbo-whisper-local-stt

This skill appears to do what it says: a local, offline-friendly STT tool using Faster‑Whisper. Before installing or invoking it, consider: - Resource & network: the first run will create a virtualenv and install packages (including PyTorch) and download model files from Hugging Face—these can be gigabytes and take time/bandwidth. - Disk footprint & location: it will write models/, logs/, and create a venv directory (the code places the venv one directory above the skill root, which may be shared across skills). If you prefer isolation, run it in a controlled environment (container / VM) or edit config to point the venv/models to locations you control. - Permissions: it uses subprocess to run pip and nvidia-smi and will execute the installed Python in the created venv; review the code if you need stricter controls. - Offline options: if you are concerned about downloads, pre-download the model and pass --model_path to avoid remote snapshot_download. If you accept automatic installs and large downloads from PyPI/Hugging Face and are comfortable giving the script filesystem write access where your skills live, this skill is coherent and usable. If not, run it in an isolated environment or modify paths before use.