Inspiration
Story telling is hard on its own. Doing it front of an audience of strangers? Impossible. More than 70% of people have public speaking nervousness, anxiety or fear. This means that a large majority of people struggle with story telling in front of a crowd. We also have also personally felt this struggle: the room spinning as you step up to the podium or stand in front of the class. As a result, we wanted to help people who struggle with public speaking prepare for their presentations and speeches better. So, we built SpeakUp and used neurology facts to help people improve their public speaking.
What it does
SpeakUp provides 4 features that help people prepare for public speaking:
- Script organizer
- Script memorizer
- Confidence and nervousness tips
- Teleprompter
How we built it
We split each of the four features into their own components. Then we began implementing the features. A google doc and a miro board were used for planning to make sure that everyone was on the same page and to make sure we were not missing an important architectural detail.
Some neurology things
(1) For the script memorizer, we used a spaced recall algorithm with flashcards. We decided to use a basic algorithm instead of FSRS because there are not a lot of flashcards. The basic algorithm shows wrong flashcards earlier and right flashcards later. (2) The tips we provide are rooted in neurology, but we also include general tips that are agreed upon by general consensus. (3) The flashcards and teleprompter have emojis because corresponding meaning and visual information improves memory recall performance.
Challenges we ran into
This was the first time any of us had used AI APIs. We were unfamiliar with how the API Key was supposed to interact with the code and which permissions to set for the API Key. Eventually we figured it out and now understand how to use OpenAI's API.
Because of the time crunch, implementing a full AUTH + DB scheme would take too much time. So, we had to find a new place to store the data. We chose to store the section and script data in the localstorage, but this was our first time interfacing with the localhost. Thus, we ran into multiple bugs in our implementation.
Forgetting to commit during the first half of the hackathon =(
Some cool things & accomplishments that we're proud of
(1) For the script organizer, the sections are generated by prompting AI. We had the AI produce a JSON object, and surprisingly, the AI is rather consistent (100% success rate). (2) We are proud of the localhost class that auto-saves any data and provides global access. (3) We are also proud of the transition of the script text bubble when switching from the "home" and "edit" tab. The two tabs create different instances of the same component, but it turns out you can apply a transition to them. (4) Also, I say that the miro board looks really nice
What we learned
We learned how to use localstorage, how to use the OpenAI API, and a couple of neurology facts that were interesting. We also gained first hand experience with the neurology fact that caffeine makes you awake; did you know that?!
What's next for SpeakUp
One idea that we wanted to do was to do dynamic teleprompting. Using a transcribing tool, we could record what the user has said and match it to sections in the script. Then, SpeakUp can display the current section's title and visual aid (emoji) and also the next section's title and visual aid.
Built With
- next.js
- openai-api
- react
- tailwind
- typescript
Log in or sign up for Devpost to join the conversation.