DATA472 - Group Project: Speech to Text
Tao Yan, Haritha Parthiban, Sarmilan Vignaraja, Shivanshu Tandon
This project aims to harness the capabilities of Whisper and Large Language Models (LLMs) to transform audio and speech into accurate transcripts. Subsequently, these transcripts will be succinctly summarized. The resulting summaries will then be systematically uploaded to an AWS cloud database, enabling efficient retrieval and search capabilities in the future.
-
Webpages are developed to realize all functionality: i) Audio to Transcript and Summary Page ii) Speech to Text Conversion Page iii) Summary Search Page
-
AI Models Whisper model and LLM model API are deployed to convert audio to transcript and make summarization.
-
Data Management For data management, a PostgreSQL database is deployed on the Relational Database Service (RDS) of Amazon Web Services (AWS).
Python: This project is built with HTML, CSS, Javascipt, Python 3.12.3
PostgreSQL: This project depends on PostgresSQL database to store data.
Clone the repository: git clone git@github.com:Apache-Hell/S3.git, cd S3
Install dependencies: pip install -r requirements.txt.
Install a local whisper model.
Input Googgle AI API key.
Start the PostgreSQL database in RDS server.
Start whisper web API by running S2T.py.
Flask: For web API
whisper: For whisper model
Error handling is implemented throughout the project, with errors being logged and, in critical cases, the process exiting with a non-zero status code.