An AI-powered audio retrieval system that identifies movies or TV shows simply by listening to a short audio clip. It transcribes audio speech to text and uses vector embeddings to perform semantic similarity matches against a subtitle database.
- Audio Transcription: Leverages OpenAI's Whisper model for robust speech-to-text conversion.
- Semantic Mapping: Converts text into meaning-based vectors using
SentenceTransformers(all-MiniLM-L6-v2). - Vector Storage: Efficiently indexes and queries thousands of dialogue chunks via
ChromaDB. - Modern UI: Built-in interactive frontend using Streamlit (Incoming).
- Language: Python
- AI Models: OpenAI Whisper, Sentence Transformers
- Database: ChromaDB (Vector Database)
- Frontend: Streamlit
- Clone the repository:
https://github.com/vivekanand85/movie-shazam-ai.git