AutoScribe

The main landing page
The dashboard with multiple videos in progress
More information about the project
The sign in / sign up page

Inspiration

We are all high school students, and in past we’ve all struggled with staying caught up to date. We’ve also all had the experience of needing to look through old work to find an answer or a piece of information, but being unable to find it. In the making of this project, we realized we could kill two birds with one stone.

What it does

AutoScribe uses machine-learning algorithms to transcribe and summarize a video's contents for easier access, all from a simple, user-friendly dashboard.

How we built it

We used Svelt and Sveltekit to make the front end with reactive javascript. We utilized a multitude of google cloud services, including Google’s Speech-to-Text, Kubernetes Engine, Cloud Storage, and Cloud Functions to automate the process from YouTube link to actionable notes.

Challenges we ran into

We had a multitude of challenges while connecting all of our modules across datacenters. A number of CORS issues, data flow problems, and syntax errors meant spending hours late at night searching through logs. This, piled on top of the fact that we were using new libraries and services that none of us had used before, meant that this project was a huge challenge.

Accomplishments that we're proud of

We’re incredibly proud that we were able to build such a complete product, since it’s common for Hackathon projects to be incomplete stubs, and we know users want to be taken care of from start to finish.

What we learned

We learned a significant amount about setting up cloud software using a multitude of services including APIs, compute engines, and Kubernetes clusters. It was also a first time for all of us using Svelte/Sveltekit, but we ended up being glad that we did.

What's next for AutoScribe

AutoScribe will strive to provide students with a complete and engaging experience that promotes learning and allows all types of students to gain knowledge. Such features could include tools for analyzing and providing in-depth help in subject matter past surface level content through specialized learning programs that lean on ML tools.