Use this Link for the product demo: https://www.loom.com/share/26e785d21f8043248e8957865e188e13.

( Katy Perry is just here for entertainment ;) (and Devpost doesn't support Loom links in their submission form :( ) )

Inspiration

Different content is suitable for different occasions. Audio works best when we're occupied like walking, driving, or going to the gym while text is preferable when we're short on time. Also, content creators tend to deliver media in one form and extending from podcasts into blogs as well as blogs into podcasts would expand a creators feasible audience.

What it does

Flow works by seamlessly translating between different medias, converting podcasts into text and into text summaries. With Flow, users can consume or deliver their favorite content in whichever media works best for them.

How we built it

We designed our MVP for a chrome extension. Our backend flask server is responsible for handling all API calls as well as serving local models (should we extend into this domain). Our frontend server is the chrome extension which on-click parses the active tab's DOM and searches and finds the .mp3 files; those .mp3 files are then translated using AssemblyAI and summarized using Cohere.

Challenges we ran into

We had difficulties with cohere's generated summaries, runtime (we had to memoize), and integration of the APIs.

Accomplishments that we're proud of

We were able to integrate multiple APIs together (Assembly, Cohere, Chrome) to deliver content to the user. We also had to carefully prompt engineer and hyperparamater tune the models powering these API engines to get good content in a reasonable amount of time.

What's next for Flow

We plan to expand Flow beyond a chrome extension and make an application that integrates with a user and content creator's natural workflow. Our vision is to deliver good, accessible content in all domains and so we plan to extend into images and videos.

Built With

Share this project:

Updates