Meme-it
An AI meme and sticker generator that turns your own photos into captioned memes and downloadable sticker packs.
Project Details
About the Project
Memes and stickers have become a big part of how we communicate online. Whether we're reacting to something funny, complaining about school or joking with friends, sometimes a meme says it better than words.
I wanted to make that more personal. Instead of always using existing templates or stickers of random people, what if we could turn our own photos and expressions into memes? That's how Meme-it started: a meme and sticker generator that's simple and fun, even for people with no editing experience.
What It Does
With Meme-it, you can:
- upload a photo and choose who appears, whether that's everyone in a group photo or just one person;
- pick a theme such as Work, Relationships, Life or Tech, or write your own;
- generate captions — the AI looks at the photo's visual context alongside your theme and suggests five meme captions you can edit;
- create a sticker — background removal turns the photo into a transparent PNG sticker with your caption; and
- build a sticker pack of 6–12 stickers to rearrange and download as a ZIP file.
How I Built It
The frontend is React, TypeScript and Vite, and the backend is Python with FastAPI, which sends the photo and theme to a multimodal model to write the captions. I refined the prompts to focus on the photo's visual context, using the theme as inspiration, so the captions are funny and relevant rather than generic jokes.
Background removal runs directly in the browser with @imgly/background-removal, and JSZip packages the sticker pack.
Project Satisfaction
I'm most proud of turning Meme-it from a simple idea into a working AI web app. The most satisfying moment was finally seeing a sticker generate on the deployed website after repeatedly troubleshooting backend crashes.
Biggest Challenge
Background removal kept crashing my backend in production: the image-processing model needed more than the 512 MB of memory my free hosting allowed. I redesigned the workflow so background removal runs in the user's browser instead, which took the load off the backend and got sticker generation working.
Key Takeaway
Building an AI application involves much more than connecting to a model. My biggest takeaway was the difference between code that works locally and code that works reliably in production.