dinscribe 0.1.1

Audio denoising and transcription pipeline using Demucs, Silero VAD, and Whisper

dinscribe audio transcription

Processes audio through a three-step pipeline to produce a transcription JSON: denoise, voice activity detection, transcribe.

Setup

python -m venv venv
source venv/Scripts/activate
pip install -r requirements.txt

Run the full pipeline

python main.py input/audio.mp3          # Run for a single file
python main.py input/                   # Run for all audio files in a folder
python main.py input/audio.mp3 -f       # Force re-run all steps

Each step checks whether its output already exists and skips it if so. Use -f to force all steps to re-run regardless, -o <output_dir> to specify a different output directory, and -c <config.yaml> to specify a different config file.

Output is written to output/<filename>/ and contains:

• <filename>_denoised.wav (vocals isolated from background noise)
• <filename>_vad.json (detected speech segment boundaries)
• <filename>_transcription.json (final transcription with timestamps)

Configuration

Edit config.yaml to adjust settings for each step. Some important options are:

• denoise.model - Demucs model for vocal isolation (default: htdemucs)
• vad.threshold - VAD speech detection sensitivity (default: 0.5)
• transcribe.model - Whisper model size tiny through large (default: base)
• transcribe.language - Transcription language code (default: en)

Other tips for best results

Add domain-specific vocabulary to vocab.txt to improve transcription accuracy on unusual words and jargon. For noisy or technical audio, set temperature: 0 to disable attempts to fallback to higher-temperature decoding, and consider filtering out any common hallucinations specific to your dataset.

Run individual steps

Each step can also be run alone:

python denoise.py audio.mp3
python vad.py audio_denoised.wav
python transcribe.py audio_denoised.wav audio_vad.json