About

Hello! This app was built using HTML, CSS, Python Flask, and JS by Navneeth Krishna. The app allows you to generate summaries for YouTube videos using OpenAI's Whisper API (speech-to-text transcription) and GPT-3.5 (summary completion).

The app interacts with a Flask server hosted on Heroku. The frontend and backend communicate via REST APIs. The launch of GPT-4 motivated me to build this product and I plan on optimizing API costs using wave processing and other techniques. In addition, I plan on skipping transcribing for videos with readily available subtitles which will further save costs.

I used GPT-3.5-turbo for this instead of GPT-4 for the former's cost-effective usage.

Be aware of the pricing of the Whisper API and gpt-3.5-turbo Chat Completion API here. In general, the Whisper API accumulates costs quickly. gpt-3.5-turbo is quite affordable.

Note that the library for downloading YouTube videos--PyTube has a bunch of errors which may lead to errors in processing.

The Backend Stack/Tech: Whisper API, GPT-3.5-turbo, Python Flask, Firebase NoSQL real-time database (Cloud Firestore), RESTful API, and pytube.

The Frontend Stack/Tech: HTML, CSS, and JS.

Screenshots