Project Overview
Over the past couple of days, I’ve been developing an AI-powered Presentation Generator to simplify creating professional slideshows from PDF documents. (I'm new to code - AI is helping with my ideas)
The Goal
My goal was simple: enable myself to upload PDFs and have draft presentations generated with minimal effort. Now, after real-world testing, I’d like to share what we learned.
How It Works
The AI Presentation Generator turns any PDF into a structured, editable presentation in five steps:
- Upload a PDF
- Specify presentation details (e.g., topic, style, presenter)
- Give it specific instructions
- AI analyzes the content and auto-generates slides
- Users can edit the slides in-browser
- Download the final PowerPoint
screen video recording of the application PDF to PPT Converter with AI
Personal Learnings
- Learning Python – Developing this project helped me pick up Python quickly. What would have taken months to learn before, AI guided me through at a faster pace.
- AI as a Tutor – AI proved to be an adaptable tutor, adjusting its conversational style to my needs, reinforcing that AI is indeed the future.
- AI’s Strengths – AI is excellent at summarization and understanding context, which was invaluable in this project.
- AI’s Limitations – I learned not to give AI fully working code to "fix" as it can sometimes break it further.
- Diversify AI Use – Relying on a single AI is not ideal. I used both Sonnet 3.5 and Gemini, which helped solve issues that Sonnet couldn’t address (though not every time).
Bugs (will fix these in the next version)
- Slide Count Discrepancy – Initially, the generator displayed 12 slides in the edit mode, while the final PowerPoint download only had 10. This discrepancy happened because I had instructed the tool to include a "Contact" slide, which was excluded from the final 10-slide limit per the original instructions.
- Missing Images – One issue we faced was that images weren't showing up in the slides. This happened because I forgot to input the API key for image generation. Once the API key was added, the image generation worked seamlessly.
Would you like to try it out or would you want me to make the code available?
[link] [comments]