Qwibo¶

Qwibo is a Windows desktop app that transcribes audio and video to text offline, using the NVIDIA NeMo Parakeet TDT 0.6B v3 model. Optional LLM summaries run locally (Qwen) or via cloud APIs.

Bilingual documentation

This site is also available in Italiano. Use the language selector in the top navigation bar.

What it does¶

Feature	Description
Transcription	Audio/video → plain text + SRT subtitles
Job queue	Multiple files processed one at a time; history and detail pages in the app
Summary	Multi-provider LLM (DeepSeek, OpenAI, local Qwen, …)
Desktop app	Electron installer — no Python, no terminal, no Docker
Privacy	Audio stays on your PC during transcription

Who this is for¶

Qwibo is built for people who already have recordings — meetings, podcasts, interviews, lectures — and want usable text without uploading files to a cloud service or paying per minute.

For self-hosted Docker on a home server or NAS, see the separate qwibo-docker repository.

Quick start¶

Download Qwibo-Setup-*.exe from GitHub Releases.
Run the installer (see SmartScreen if Windows blocks it).
Complete the first-run setup (~4–5 GB one-time model download).
Upload a file and click Queue transcription.

Details: Download & install · First run

Hardware at a glance¶

Resource	Minimum	Recommended
Windows	10/11 64-bit	11
RAM (transcription only)	8 GB	16 GB
RAM (local Qwen summary)	16 GB	32 GB
Free disk	8 GB	12 GB
GPU	Not required	NVIDIA optional (future)

Full details: System requirements

Documentation map¶

Section	Content
Getting started	Install, first run, models
User guide	UI, queue, summaries, formats
Troubleshooting	Common errors
Reference	FAQ, licenses

Build documentation locally¶

pip install -r docs/requirements.txt
mkdocs serve

Preview at http://127.0.0.1:8000 — publish with scripts\publish_docs.bat (or python scripts/publish_docs.py), then git push in the sibling repo qwibo.github.io. No GitHub Actions.