pszemraj/vid2cleantxt

Leverages OpenAI's Whisper model within a multi-stage NLP pipeline that chunks video into manageable audio segments, transcribes via transformer inference, then applies spell-correction (Neuspell), sentence boundary detection (pySBD), and keyword extraction (YAKE) to produce publication-ready text. Integrates with Hugging Face's transformers library and supports both CLI and Python API interfaces, with optional Google Colab deployment for GPU-accelerated processing without local hardware requirements.

220 stars and 19 monthly downloads. No commits in the last 6 months. Available on PyPI.