After trying in vain to find the source of a number of videos that we came across while on social media, we realised that you cannot use a traditional search engine to find a video based on the content contained within it.

What it does

VideoFind is a very early version of a new search engine that utilises a new way to find and catalogue videos based on the visual and audio content that is stored within.

This version of VideoFind allows users to upload videos to it after which the video is hashed, the conversations within are collected and the relevant topics within these conversations are saved.

This allows for users to search for videos based on the words spoken in the video as well as through the use of screenshots.

How we built it

VideoFind was a complex system to build and it required the training of neural network models that allow for videos to be hashed after which we utilised google cloud speech to text APIs to convert the conversations in the uploaded videos to textual data. APIs were then used to extract the relevant conversation topics from the transcribed conversation.

Challenges we ran into

Working with video files was difficult. This compounded with the compute requirements needed to transcribe audio data from videos made the implementation of an effective video search system difficult.

Accomplishments that we're proud of

We managed to build a new way to find videos online which was awesome.

What's next for VideoFind

We believe VideoFind will be the best way to find videos online and we will be adding new functionality to VideoFind that allows for videos to be searched for using finer details such as the faces of those in the video as well as the emotions of the people within the video.

Share this project: