ReeLearn

Reels that match what you want to learn.

Comment

Inspiration

Short videos are an easy way to learn — tech, sports, or just random stuff — when the feed lines up with what you care about. Most of the time it doesn’t, so you scroll past junk instead of actually picking up something useful. ReeLearn is our hack on that idea: keep the learning, lose the noise by steering Reels toward topics you choose.

What it does

ReeLearn is a Chrome extension that helps you learn from Instagram Reels on your terms.

  • Captures reel video while you browse (what’s on screen and queued clips).
  • Sends each clip to the server; Gemini returns a short summary of what the video is about.
  • You define topics in a run (e.g. tech, a sport, a hobby). The API compares the summary to those topics and returns match plus per-topic results so the extension can stay on useful reels or move on faster.
  • You can save reels you care about to the server for replays!

The backend supports Gemini via API key or a Vertex AI + cloud storage pipeline so your team can demo locally or on GCP.

How we built it

  • Frontend (the extension):
    This is the part that lives in your browser — the popup, the scripts on Instagram, and the glue that pulls the reel and sends it to our app. It’s the side you actually touch.

  • Backend (the server):
    This is where the video gets uploaded, summarized, and matched to the topics you set. It’s also what stores your runs and any reels you choose to save.

  • Database:
    A simple place to remember things between sessions — which run you’re on, what topics you care about, and saved clips if you keep them.

  • AI:
    We use Google’s Gemini to watch the clip (as a file) and describe it in text, then we compare that to your topics so the extension can decide what to do next.

Challenges we ran into

  • DEPENDENCY HELL! that one is the main one... our code wasn't working the same in our laptops, but we figured it out!
  • Compressing the video files was also a really big challenge we had

Accomplishments that we're proud of

We’re really proud that we got the full cloud path working end-to-end. We set up a Google Cloud project with Cloud Storage so our videos land in a proper bucket, and Gemini runs on that setup. It wasn’t just “call an API from a laptop”, we made the upload → storage → LLM model flow. Getting that pipeline clean was a real win.

What we learned

  • Gemini & Google Cloud
    Most “Gemini failures” were really uploads, buckets, and permissions problems WE generated... Once GCS + Vertex were wired right, Gemini gave summaries we could trust.

  • Hackathon scope
    If capture → understand → match works, you’re done, the rest is polish.

  • Ethics
    Tweaking how someone scrolls bumps into ToS and privacy — fine for right now, but worth a real conversation later :)

What's next for ReeLearn

  • Accounts & auth
    A real user system (sign-in, sessions, per-user data) so runs, topics, and saved reels belong to you!

  • Smarter sessions
    Cache reels we’ve already processed and explore async jobs if uploads or model calls start to lag.

  • Your library
    Turn saved reels into playlists you can replay, organize, and maybe export. Powered by per-user storage once auth exists.

  • Better UX
    Surface why something matched, make topics easy to edit, and make skip vs stay feel clear and intentional.

Built With

Share this project:

Updates