Pixel to Plate

Inspiration

1. My TikTok saved tab is basically a museum of delusion. It’s full of recipes I was convinced I’d cook, and then never did, mostly because of friction. The recipe is trapped inside a 45-second video where the ingredients flash on screen for 0.2 seconds. Different creators share the recipes in different formats. Vague captions and fluffy descriptions mean I've wasted sooo many minutes rewatching videos.

2. Most recipe apps are overloaded with feature bloat, when what people really need is simplicity. I wanted Pixel to Plate to be centered around doing one job, around how people realistically browse TikTok.

What it does

Pixel to Plate works with videos you share from TikTok, Instagram, and YouTube. Depending on how much detail the creator actually includes, it pulls recipe info from a mix of:

• the video description (easy)
• audio transcription
• on-screen text and captions

The best part is that it works directly from the source app. You just hit Share, tap Pixel to Plate, and the import runs in the background. Because let’s be honest, maybe youre viewing tiktoks on the train! You're not jumping into the kitchen immediately. Makes so much more sense than 'pasting a link' each time.

When you’re ready, the recipe is waiting in your library, and you can organize, filter, and save everything in a way that actually works for you.

How we built it

Vibe coded with tonnnesss of doc reading.

Challenges we ran into

• Recipe videos are not structured, they’re chaos + vibes
• Extraction takes time, so I learned very quickly that this had to be done in the background
• I hit the classic “everything saves globally for all users” bug and had to fix it fast 😭
• I am not a developer! Docs gave me a headache.

Accomplishments that we're proud of

• Building something that I'm actually using -Getting more confident using the terminal -Submitting to iOS and TestFlight (all brand new to me) -Learning how to prompt for things outside my comfort zone, especially writing clear specs for backend and data models

What we learned

1) Non engineers can build now! I already knew this but being able to use Chatgpt to learn how to get things done step by step and actually build a working app is proof. 2) Simplicity wins - there is so much feature bloat in existing recipie/food apps, it doesnt make sense to try and do everything. Also think about user flow. eg. ' Sending grocery lists to a shop isn't realistic, thats not always how people shop.

What's next for Pixel to Plate

• Faster imports and better performance in general to make it undeniably sticky.
• “Cooked it” tracking + streaks for personal motivation
• Better extraction methods, especially for step by step thumbnails,

Built With

• apple
• drizzle
• express.js
• openai
• orm)
• postgresql
• replit
• revenuecat
• tanstack
• typescript