soufianesejjari/darija-wit-ai-audio-to-srt-transcriber

Convert audio or video into accurate SRT subtitles using Wit.ai and webrtcvad. Supports Arabic & Darija (Moroccan Arabic), with automatic speech segmentation, proper timestamps, and JSON export. Perfect for generating captions, transcripts, and multilingual subtitle workflows. # Wit.ai SRT Transcriber This project provides a command-line tool to transcribe audio/video into **SRT subtitles** using Wit.ai. ## Features - Converts any audio/video file to 16kHz WAV (via ffmpeg) - Uses webrtcvad for voice activity detection - Splits audio into speech segments and sends them to Wit.ai - Outputs both `.srt` subtitle file and `.json` log ## Requirements - Python 3.8+ - ffmpeg installed and available in PATH Install dependencies: ```bash pip install -r requirements.txt ``` ## Usage 1. Set your Wit.ai server token in environment: ```bash cp .env.example .env # edit .env with your Wit token export $(cat .env | xargs) # load into environment ``` 2. Run transcription: ```bash python src/main.py path/to/audio.wav --srt out.srt --json out.json ``` ## Env vars - `WIT_TOKEN` = "Bearer YOUR_WIT_SERVER_TOKEN" ## Output - `*_wit.srt`: subtitle file with proper timestamps - `*_wit.json`: raw data (segment text + Wit response)