The Problem

With a growing amount of video content, it becomes increasingly harder to search within those videos. In addition, many of us want to only see particular parts of the video but don't want to rely on someone's word of mouth to show them a particular part. Many longform videos are poorly annotated, or don't have easily indexable captions. So we find it is really hard to find

Specific jokes in stand-up comedy routines

Emilio Estevez jokes in HackCon videos

Example problems within posted lectures

Checking for plagiarism in public addresses

Step #7 in a DIY project

The shortcomings of unikernals in a 40 minute talk about unikernals

The Solution

We decided to give people the power to search within videos. Enter a topic or a quote and get back the exact (or closest to exact) phrases that were said. So a frustrating search of scrubbing within the video is no longer. We return the exact time where the quote is being said, along with a video embedded with the timestamp loaded (if it is a YouTube video)

It's Command -F / Ctrl - F for videos

Notable Technologies Used

We used React + React-YouTube for our frontend

Java + AWS Lambda for our backend

SASS for great design

Technical Nitty-gritty

Anton Relin & Joe Azar-Williams developed a back-end on Java interacting with a AWS Lambda instance. Upon receiving a query, the server will interact with the YouTube API to get a list of relevant videos within the field. Then we take the videos, parse the audio though a cough secret internal caption API that YouTube once had public support for but then hid cough . Once we get that data, using a search library imported into the instance searches for relevance and matches using the fuzzy method. The server then gives the result through a RESTful interface. (However, the rest of the server uses Lambda for API interactions.)

This is where Abhinav Suri & Angelo Saraceno come in, using React, a JavaScript front-end library - we listen to a POST request from the API we have. Once that is in, we have the videos display and parse the JSON to get the desired results on the page. Any animations you see are pure SCSS - for a better UX, we decided to also place multiple timestamps.

What we learned

The You-Tube API is the worst API, and we were able to get out of our comfort zone. This was Angelo's first time working with a team outside of FIU. Anton and Joe became more aquatinted with AWS Lambda. Abhinav usually doesn't tough front-end.

What's next for Relaudio

Add podcasts, bookmarks, auth, users, and show Chris Sacca if we can with the Start-Up Podcast Episode 1 as our demo example.

Built With

Share this project: