This project provides a deep learning-based system to generate text transcripts from YouTube videos using Automatic Speech Recognition (ASR). By leveraging Hugging Face's Wav2Vec2 model, the project converts spoken words from video into accurate text. It focuses on fine-tuning the ASR model to handle a variety of accents, languages, and noisy environments effectively.
- Audio Extraction: Downloads and extracts audio from YouTube videos.
- Speech-to-Text: Transforms the audio into text using a fine-tuned Wav2Vec2 model.
- Customization: The model can be fine-tuned for improved accuracy on specific datasets or types of speech.
- Performance: Evaluates accuracy and speed of transcription across different audio qualities.