Inspiration

We're cofounders of Upriver.ai, an API that identifies what's influencing customer demand (e.g., trends, products, creators). We enrich advertising and sales workflows so teams have confidence in the quality and relevance of execution.

Website: https://upriver.ai Docs: https://docs.upriver.ai

Team

Our team has extensive experience building consumer ad platforms.

Lulu (CEO) - prev design, product, eng @ Twitch, ex-Stanford Joe (CTO) - prev eng @ Instagram, Google, Spotify. Led implementation of IG Live, Reels, and video calling

About Swipe (problem/solution/business impact)

Sponsorship ads are a key part of the advertising ecosystem. Brands partner with creators to reach their audiences, and the best creators translate brand value into something their followers care about.

The problem is that it takes creative teams extensive resources to identify and analyze sponsorship ads.

Swipe helps creative teams identify, classify, and contextualize sponsorship ads on YouTube. With Swipe, teams can understand how consumers are hearing about specific products and from whom. These insights allow them to get into the consumer's mindset and ultimately make more relevant ads that perform.

How we built it

  • /sponsors and /sponsorships endpoints from Upriver.ai for identifying YouTube sponsorships
  • python, FastAPI,TwelveLabs for video understanding
  • next.js, tailwindcss, shadcn for UI

Challenges we ran into

  • On-screen element extraction, for getting timestamps of a brand logo appearance
  • Ways to download, trim and upload a youtube video quickly
  • The distinction between the two TwelveLabs models; ended up having to reindex our collection

Accomplishments that we're proud of

Getting video understanding up and running with TwelveLabs.

What we learned

Touching grass is really helpful mid-code sprint.

What's next for Swipe

We'll incorporate video understanding aspect into our api.

Built With

  • next.js
  • tailwind
  • upriver.ai
Share this project:

Updates