Inspiration

I have struggled to find specific audio files on my phone, whether it's a birthday wish from a loved one, a reminder I set for myself, a WhatsApp Audio message, etc. I have often spent hours searching for the exact audio click. Given the exploding number of personal audio/ video files, it's super challenging to organize them and find them when you need them. Google Photos does a great job at organizing photos based on the people located and helps us identify them easily. I want MediaBox to do the same for audio/video content based on the audio content

What it does

Mediabox transcribes your audio content with the power of AssemblyAI APIs and indexes them via a search engine so that you can easily search for your audio files. Mediabox can also play your audio content directly and also shows you your audio content transcribed

How we built it

Mediabox is built on a Serverless Architecture on AWS. The front end is built using Angular and the back end is built via Node. For the database, it used Dynamodb and stores files in S3. Mediabox uses Algolia as its search engine and uses AssemlyAI APIs to transcribe the audio content.

Challenges we ran into

I wasn't able to get super accurate transcription on Audio files. I also wanted to build a separate classifier for organizing music files separately but didn't get enough time to do the same.

Accomplishments that we're proud of

I am proud that I was able to build a fully functional demo within 2 days with a robust and scalable architecture. The web application is hosted and I have shared a CloudFront URL. Anyone can signup and create their own account on MediaBox and make their audio files searchable.

What we learned

While I worked extensively in AI during my college days(5 years back), I stopped working on the same as I began a career in software engineering. I am astonished by the developments in AI during this period and the accuracy of the new AI models. Working on this hackathon really motivated me to learn, research, and built more products using AI.

What's next for Mediabox

From a practical user perspective, I believe a mobile application is essential as it allows users to automatically upload their media content directly without them manually uploading the same. From a functionality standpoint, I would like to build classification based on the type of audio (eg: music vs non-music) and also categorize the files based on the content, for example, automatically building a to-do list based on audio messages say from your loved one and reminding you about the same.

Built With

Share this project:

Updates