AWS Builder Center
Weekend Creative Challenge: Meme-Remix Booth

Weekend Creative Challenge: Meme-Remix Booth

Remix Booth is a browser-based meme maker: drop in a photo, snap one live with your camera, or drag one over, and the app turns it into a shareable meme in seconds

Cloud, Data | AWS Student Builder Group Leader

Vision & what the app does

Remix Booth is a browser-based meme maker: drop in a photo, snap one live with your camera, or drag one over, and the app turns it into a shareable meme in seconds. Type a top and bottom caption and it renders in the classic white-with-black-outline meme style, auto-wrapping to fit. If you're out of ideas, a "🎲 Random caption" button pulls from a small built-in bank of deliberately dumb captions to get you unstuck. From there you can push the image through a filter — grayscale, sepia, invert, a saturated "hot pink" look, or a genuinely glitchy RGB-channel-split effect with sliced horizontal offsets — tilt it a few degrees for that off-kilter meme energy, and download the result as a PNG.
The creative output is intentionally low-stakes and fast: the whole loop from "open the page" to "download a finished meme" takes under thirty seconds. That mattered more to me than feature count. Memes are a disposable, funny medium, and the tool should feel disposable and funny too, not like a photo editor with a learning curve.

How you built it

The entire app is a single static HTML file — no framework, no build step, no backend — using the Canvas 2D API for image compositing and text rendering, the FileReader API for uploads, and  getUserMedia  for live camera capture.
Key decisions and the reasoning behind them:
  • Canvas over CSS overlays for text. Early versions tried layering HTML text on top of an  <img>  with absolute positioning. That fell apart the moment I needed a downloadable PNG — you can't flatten DOM elements into an image without extra libraries. Moving everything onto a single canvas (image, filters, and text all drawn in the same 2D context) meant  canvas.toDataURL()  "just worked" for export, and it's also what let filters and text interact correctly when the canvas is tilted.
  • Cover-fit cropping. Uploaded photos come in all aspect ratios, but the canvas is a fixed square. I scale each image up to cover the square (rather than fit, which would leave letterboxed bars) and center-crop it — the same logic Instagram-style square crops use.
  • Camera capture was the trickiest addition. Getting a live  <video>  preview to feel like a real camera booth meant mirroring the feed horizontally ( transform: scaleX(-1) ) so it behaves like a mirror, then re-mirroring the captured frame during the canvas draw so the saved image matches what the user saw, not a flipped version of it. I also had to crop the raw camera frame to a centered square before drawing, since webcams don't output square video natively.
  • Glitch filter without a library. Rather than reach for a WebGL shader, I manipulated raw pixel data directly via  getImageData / putImageData : shifting the red and blue channels a few pixels in opposite directions, plus randomly offsetting a handful of horizontal slices. It's a handful of lines of vanilla JS and looks far more broken (in a good way) than a simple CSS filter could.
  • A real constraint I hit late: camera access requires a secure context (HTTPS or localhost). A plain S3 static website endpoint serves over HTTP only, which silently blocks  getUserMedia  in the browser. That pushed the architecture from "just S3" to "S3 behind CloudFront."

AWS services used / architecture overview

Browser (upload / camera / canvas render)
│ HTTPS
â–¼
Amazon CloudFront (CDN + HTTPS termination)
│
â–¼
Amazon S3 (static website hosting, origin bucket)
  • Amazon S3 stores and serves the single  index.html  file as a static website — the origin for everything.
  • Amazon CloudFront sits in front of the bucket as a CDN, and critically, terminates HTTPS with a free CloudFront domain certificate. This is what makes camera capture work at all in a real browser, plus it gets edge caching and lower latency for free as a side effect.
No compute, database, or API layer is needed — the app is fully client-side, so the "backend" is really just "serve these static bytes over HTTPS reliably," which S3 + CloudFront does within AWS Free Tier limits for a project this size.

What you learned

The biggest lesson wasn't about memes or canvas rendering — it was a browser security detail I'd always half-known but never hit in practice:  getUserMedia  categorically refuses to run over plain HTTP. I originally deployed with S3 static website hosting alone and the camera button silently failed with a permissions error that had nothing to do with actual camera permissions. Adding CloudFront in front of the bucket to get HTTPS was a small architectural change but a real "oh, that's why" moment, and it's a good reminder that "static site" doesn't mean "no infrastructure decisions" — TLS termination is still a decision.
On the AWS side specifically, this was my first time wiring CloudFront to an S3 origin from scratch rather than through a higher-level tool like Amplify, and it clarified how origin access control, cache behaviors, and the distribution's default root object all fit together for a dead-simple single-page static app.

Link to app or repo

Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article