AI prepTalk

Inspiration Preparing a presentation can be stressful, especially when the Q&A section is unpredictable. Even if someone has prepared their slides carefully, they may not know what questions the audience will ask or whether their answers are clear and relevant. We wanted to build a simple tool that helps people practice this part of a presentation before presenting in real life. What it does PrepTalk is an AI-powered presentation practice coach. Users enter their presentation topic and choose their target audience, such as classmates, professors, or judges. PrepTalk then uses Google Gemini to generate three realistic questions that this audience might ask. The user selects one of the questions and types their answer. PrepTalk analyzes the response and provides structured feedback, including:

• a clarity score
• a relevance score
• one key strength
• one improvement suggestion
• an improved sample answer The goal is to help users become more prepared and confident when answering questions after a presentation. How we built it We built PrepTalk using HTML, CSS, and JavaScript for the frontend, with Node.js and Express for the backend. Google Gemini is used to generate audience-specific presentation questions and evaluate the user's written answers. We divided the project into four main parts:
• frontend and user interface
• AI question generation
• answer feedback logic
• integration and testing The frontend sends user input to the backend through API requests. The backend then communicates with Gemini and returns the generated questions and feedback to the interface. We also used environment variables to keep the Gemini API key private and separate from the submitted source code. Challenges we ran into One of our biggest challenges was connecting the different parts of the project into one working application. We had to make sure the frontend and backend communicated correctly, that the question-generation and feedback modules worked together, and that Gemini returned responses in a consistent format that could be displayed properly. We also ran into setup and debugging issues related to Node.js, npm, API requests, model responses, and application integration. Because we were working within a limited hackathon timeframe, testing and debugging the complete application flow became especially important. Accomplishments that we're proud of We are proud that we were able to build a working prototype within the hackathon time limit. We successfully created a complete flow where a user can:
• enter a presentation topic
• choose an audience
• generate realistic AI-powered questions
• select a question
• submit an answer
• receive structured feedback We also successfully combined separate frontend, AI, feedback, and backend components into one working application. What we learned Through this project, we gained experience with:
• frontend and backend integration
• JavaScript
• Node.js and Express
• working with APIs
• Google Gemini
• handling JSON responses
• environment variables and API key security
• debugging and testing
• GitHub and project collaboration We also learned how important communication, task division, and integration are when building a project collaboratively under time pressure. What's next for PrepTalk In the future, we would like to expand PrepTalk with features such as:
• voice recording
• speech-to-text
• AI-generated voice feedback
• speaking-speed analysis
• presentation history
• progress tracking
• more detailed scoring
• support for different presentation styles and difficulty levels
• personalized coaching based on previous responses Our goal is to make PrepTalk a more complete presentation practice tool that helps users prepare not only for what they want to say, but also for the questions that may come afterward.

Built With

• ai
• api
• css3
• express.js
• gemini
• generative
• google
• html5
• javascript
• json
• node.js
• rest