Models¶
Qwibo uses one required model for transcription and optional models for LLM summary. End users download these via the first-run wizard — no manual scripts.
ASR model (required)¶
| Property | Value |
|---|---|
| Hugging Face | nvidia/parakeet-tdt-0.6b-v3 |
| Local file | %APPDATA%\qwibo-desktop\models\parakeet-tdt-0.6b-v3.nemo |
| Size | ~2.5 GB |
| Engine | NVIDIA NeMo |
Optimized for Italian; other languages work with varying quality.
Cloud LLM summary (optional)¶
No model on disk. Requires an API key for the chosen provider.
| Provider | Configure in app |
|---|---|
| DeepSeek | Summary settings |
| OpenAI | Summary settings |
| Gemini | Summary settings |
| Claude | Summary settings |
| Kimi | Summary settings |
Keys are stored in %APPDATA%\qwibo-desktop\data\.secrets\summary_keys.json.
Local LLM summary — Qwen (optional)¶
| Property | Value |
|---|---|
| Model | Qwen2.5-3B-Instruct (GGUF Q4) |
| Path | models\qwen2.5-3b-instruct\*.gguf |
| Size | ~2 GB |
| RAM | ≥ 16 GB system RAM |
| Engine | llama.cpp (CPU) |
Downloaded automatically during first-run setup when RAM threshold is met.
Summary table¶
| Component | Location | Download | Network at runtime |
|---|---|---|---|
| Parakeet ASR | models\*.nemo |
Setup wizard | No |
| Cloud summary | — | API key only | Yes |
| Local Qwen | models\qwen2.5-3b-instruct\ |
Setup wizard (if RAM OK) | No |
Reinstall / clean slate¶
Uninstalling Qwibo does not delete models. To force re-download:
- Quit Qwibo from the tray.
- Delete
%APPDATA%\qwibo-desktop\models\. - Restart Qwibo — setup runs again.
Licenses¶
- Parakeet: NVIDIA / NeMo terms
- Qwen: Alibaba model license — see Hugging Face
- Qwibo app: proprietary — see Licenses
Developers maintaining a dev build use scripts/download_model.py in the repo; that is not part of the retail installer flow.