An interactive karaoke system with speech recognition and real-time word matching.
- Enhanced speech recognition with robust error handling and auto-restart
- Captures 15 words of context around target words for improved accuracy
- Advanced scoring using Jaro-Winkler (≥0.8) and Trigram (≥0.4) similarity algorithms
- Automatic microphone permission management
- Real-time display of target words, spoken words, and matching scores
- Smooth auto-scrolling to keep current sentence always visible during playback
- Visual highlighting of active listening windows and matched words
- Sentence-aware scrolling that shows complete phrases, not just individual words
- Uses total song length from Whisper transcription for accurate timing
- Song progress bar and timestamps based on actual audio duration
- Supports outro periods beyond lyrics for complete song experience
- Multiple themed songs with custom styling
- Responsive design with smooth animations and transitions
- Robust error handling for audio playback and microphone access
- Visual feedback for word matches and listening states
- Write a song, or let GPT create a short poem.
- Use AI to create a song, based on the poem.
- Extract the timestamped lyrics from the song.
- Extract the music only version of the song.
- Convert the lyrics to lyrics data using the Python parser