Inspiration
For remote workers, effective online presentation skills are a necessity to convey ideas and advance ones career.
As an introvert, I struggle presenting my concepts in online meetings and lack opportunities to get feedback and improve my skills.
Advances in AI make it possible to practice presentations, get feedback on clarity, visualize my use of fillers or other poor habit, and become a better presenter
What it does
Users record their presentations. These are transcribed and processed by Assembly AI and then displayed back to the user. The user can read through the transcript and click any portion to hear it again
We also provide visualizations of problem areas in the text
Finally we use GPT-3 to suggest other ways to improve the text
How we built it
The audio recording and assembly AI were built in python. The user interface for visualizing is built in React/html5/css
Challenges we ran into
I spent a lot of time trying to find a Grammarly API that could provide improvements to the sentence structure and make suggestions. Unfortunately, I did not find any that were python based and performed well enough for integration. Therefore, I pivoted to using GPT-3
I decided to record the audio and save the data locally to save on time - a cloud based solution would be preferable, but would also need authentication to be usable, and that was outside of this one person team.
Accomplishments that we're proud of
I'm excited by the initial results of recording and being trained in better presentation. The ability to skip to portions that need improvement are really good.
AssemblyAI really outperformed expectations. I think that having a learning system to look at repeated words and other poor habits will be easy to implement in the future.
What we learned
I learned a bit about python audio and learned a lot about the AssemblyAI api. One of the main reasons I wanted to do the hackathon was to improve on my ability to cut scope and finish a project and I think that I learned a bit that even when tough decisions are made to cut scope, the product can still be pretty cool.
What's next for Prezo Coach
Well cloudbased, user authentication for one
Definitely will start to build out more of the coaching models that will use some proprietary machine learning based to get presentation specific concepts.
I also want to turn the visualizer into an editor where the user can change the structure and re-write their presentation.
Finally, I want to integrate AI to do suggested text and synopsis.
Log in or sign up for Devpost to join the conversation.