FoodProcessor

Inspiration

I love to cook and watch youtube cooking shows all the time. I am also lazy an do not want to watch the full video for instructions and ingredients. So I built this app to save time and get started with cooking faster.

What it does

Takes a YouTube url for a cooking video
Extracts ingredients and cooking instructions, orders them in a nice list
Created share-able shopping lists based on the ingredients
Creates recipe cards with ingredients and instructions (and a nice image)
Offers swap ingredients feature to change individual item to suit different diets like vegan, keto, dairy-free etc.

How we built it

I built is myself, 2 projects backend and frontend. I use GCS backend since it is easier to use than AWS. Neon.tech for db, used Claude code for coding, my brain for architecture :-)

Challenges we ran into

Time, since I am building a few other projects. Separation of free and premium content, were free users get a limited list of ingredients and cooking instructions took some time. Apart from that, it is only work.

Accomplishments that we're proud of

The system uses a tiered escalation pipeline. Each video enters at Tier 0 and only advances to a more expensive tier if the previous tier fails or produces low-confidence results.

Tiered Escalation Pipeline

YouTube URL ↓ Tier 0 — Metadata Scrape (cost: ~0 USD/video) • Parse description, comments, linked blog/recipe sites If confidence < threshold → next tier ↓ Tier 1 — Transcript + LLM (cost: ~0.001 USD/video) • Fetch YouTube captions → LLM ingredient extraction If confidence < threshold → next tier ↓ Tier 2 — Video Analysis (cost: ~0.01–0.03 USD/video) • Gemini 2.5 Flash processes video (visual + audio) If confidence < threshold → next tier ↓ Tier 3 — Deep Scan (cost: ~0.36 USD/video) • Gemini 2.5 Pro or multi-model cross-validationhttps://youtube.com/shorts/odbWZrSx1R4

Together these tiers can handle about 90% of youtube cooking videos.

What we learned

Building with expo is easy but expensive.

What's next for FoodProcessor

There will be a Tier 4 layer for more complex videos with frame analysis.
Gamification - daily/weekly challenges, earn stars and rewards and push messaging to improve engagement.
Support for additional video platforms.
PDF exports.
Nutrition value calculation
Global pantry - show what you already have at home