xkmato/luganda-transcriber

# Luganda Speech Recognition & Transcription System A high-performance Automatic Speech Recognition (ASR) application designed for transcribing **Luganda** and other Ugandan languages. Powered by the state-of-the-art **Sunbird SALT Whisper Large v2** model (`Sunbird/asr-whisper-large-v2-salt`), this project provides a modern web interface with real-time progress streaming, a feature-rich CLI tool, and flexible local/offline execution options. --- ## ⚡ Quick Setup for Non-Developers (AI Agent Method) If you are not a developer or prefer not to configure command-line environments manually, you can use an AI Coding Agent (such as **Google's Antigravity**) to set up and start the application for you: 1. **Install an AI Coding Agent**: Install an AI coding assistant like **Google Antigravity**. 2. **Copy the Repository Link**: Copy the URL of this repository: ```text github.com ``` 3. **Ask the Agent to Run the Server**: Open your AI agent, paste the repository link, and ask: > *"Please clone this repo `github.com` and run the server for me."* 4. **Access the App**: The agent will automatically set up the virtual environment, install dependencies, download required models, and launch the server. Once running, open your web browser and go to `localhost`. --- ## Key Features - **Multi-Language Support**: Optimized for Ugandan languages: - **Luganda** (`lug`) - **Acholi** (`ach`) - **Lugbara** (`lgg`) - **Runyankole** (`nyn`) - **Ateso** (`teo`) - **Ugandan English** (`eng`) - **Real-Time Progress Streaming**: Web application uses Server-Sent Events (SSE) to display chunk-by-chunk progress during transcription. - **Multiple Output Formats**: Export transcripts as Plain Text (`.txt`), SubRip Subtitles (`.srt`), WebVTT Subtitles (`.vtt`), or detailed `JSON`. - **Command-Line Tool (`cli.py`)**: Easily transcribe local audio files from the shell for batch jobs or automated w …