Inspiration

The idea came from the presentation of DRW during the opening ceremony. The presenter was reading a part of his speech on his phone and had to scroll down manually through the speech. I thought it’ll be great to have a system that follows up where you at, and gives you a feedback about how fast you’re reading.

What it does

It does many important things that are essentials for a good speech :

  • It listens to you during your speech through the microphone and auto-scrolls step-by-step so you don’t have to manipulate your computer/phone during your presentation.
  • It highlights your progress, to let you know how far you’re in the presentation.
  • It gives you real-time feedback about your pace : if you’re on time, slow or too fast.

How we built it

To build this tool, we had to break it in very smalls step-by-step process :

  • Access to the microphone
  • Create chunks based on the given text
  • Get the transcript received from the microphone
  • Try to match the transcript with chunks of the given speech
  • Analyse using AI if the sentence based on the context matches the chunk.
  • Scroll programmatically to that chunk

Challenges we ran into

The most challenging part of the project was finding the right way to create those chunks, determine when should we cut a part or not. We also had some issues on matching the chunks with the transcript before finding a js library (fuse) that does it pretty well.

Accomplishments that we're proud of

Did it alone in 24 hours, there was so many challenges in that very simple app. From seperating in chunks, to compare with transcript and many others small issues.

What we learned

Even the simplest ideas can be pretty hard to execute.

What's next for Conupitch

I'll try to continue the project on my own and maybe develop it better, enhance performance and potentially try to sell it as a SAAS

Built With

  • fuse.js
  • nextjs
  • openai
Share this project:

Updates