Overview¶
Qwibo turns audio and video into text you can search, edit, and reuse — on your own Windows PC.
How it works¶
Installer (Electron)
└── Embedded Python backend (FastAPI + worker)
├── Parakeet ASR → trascrizione.txt + sottotitoli.srt
└── Optional LLM → riassunto.txt
You interact through a desktop window (not a browser tab you manage yourself). The app starts the backend automatically and keeps it running in the tray while Qwibo is open.
Design principles¶
- Local transcription — after the one-time model download, audio is not sent to the cloud for ASR.
- Offline models — Parakeet and optional Qwen GGUF live under your user profile.
- No overwrites — each job gets its own timestamped folder.
- One job at a time — the worker processes the queue sequentially to limit RAM use.
Data location¶
All user data is under Electron userData (typically):
%APPDATA%\qwibo-desktop\
├── data\ ← jobs, input, output, secrets
├── models\ ← Parakeet .nemo + optional Qwen GGUF
├── logs\ ← diagnostic logs (tray → Open log folder)
└── cache\ ← internal caches (numba, matplotlib)
Open logs
Right-click the Qwibo tray icon → Open log folder if something fails during setup or transcription.
Alpha status¶
Current channel: 0.1.0-alpha — Windows installer, unsigned (SmartScreen warning). Mac and code signing are planned.