Skip to content

Sbobinator

Overview

sbobinator/sbobinator

Overview¶

Sbobinator is designed for two usage modes:

1. Development and desktop use (native Python)¶

Windows, Linux, or macOS with Python 3.12+
Install with scripts/install_local.py
Models in models/ at the project root
Data in data/input/ and data/output/
Start with start.bat (Windows) or sbobina ui

Best for: testing, personal use, benchmarks on your own machine.

2. Production (Docker)¶

Linux image with dependencies and ASR models baked into the build
Only the data/ folder mounted from the host (input + output)
No download on first container start

Best for: mini PCs, servers, repeatable deployment.

Typical user flow¶

flowchart LR
    A[Audio/video file] --> B[UI upload or CLI]
    B --> C[SQLite queue]
    C --> D[Worker]
    D --> E[NeMo transcription]
    E --> F[Export TXT/SRT]
    F --> G{Summary?}
    G -->|Yes| H[LLM summary]
    G -->|No| I[Job folder]
    H --> I

Current version¶

0.3.0 — SQLite job queue, worker in a separate process, job folders YYYYMMDD_HHMMSS_filename.

Next steps¶