Inspiration

I don't like reading long PDFs. Honestly, I don't even like short ones. I figured it would be way more fun if the PDF could just talk to me instead. So it did! I made it talk - not directly to me, but to another person - and now I can keenly eavesdrop on their conversation.

What it does

Buddy-please-study is the only AI agent you’ll ever need to study. It takes PDFs, understands the concepts, generates a dialogue between two people discussing them, and then speaks it out loud. And that’s not all — there are modes! You can choose between fun options like: "Teach me like I'm 5", "Yell at me", "Brain rot" and more. Nothing sticks in your brain better than a lively (and educational) chat! You can also input your own custom prompts instead of a PDF.

How I built it

I used Groq (the real MVP of my project) for text-to-speech. Gemini API for AI responses. Angular for frontend and Flask for backend. I used pdfplumber to extract text from PDFs, Gemini API to convert the text into a conversation, Groq to convert the text conversation into speech. I used base64 encoding to map audio clips instead of saving audio files, which kept things lightweight.

Challenges I ran into

Classic CORS errors, python3 issues on Mac, and angular router errors. Thanks to the mentors I was able to pull through!

What's next for buddy please study

Optional follow up quiz after a lesson, which would be interactive (taking in user speech as answers).

Share this project:

Updates