← (hackathons)
ai-recipe-feed.jpg
FoodieBag – AI Recipe Videos
TikTok-style iOS feed of AI-generated recipe videos, with a recommender and snap-a-dish recipe recognition.
Context
FoodieBag was built at hackaTUM 2023 (November 2023) with teammates Bethany Wong, Mostafa Elhayani and aspiforgreat.
Pitch: Not everyone is born with a chef's hat — but what if cooking were as easy as watching a personalised cooking show?
What we built
Two things in one weekend:
- An AI video generator that creates personalised recipe videos (AI-generated presenter videos, including via Synthesia).
- An iOS app with a TikTok-style "For You" feed of AI-generated recipe videos, each with nutritional facts and ingredients.
- A recommender system tailors the feed to your interests and dietary restrictions.
- Snap a photo of any dish and the app reveals the recipe and ingredients and adapts it to your preferences and diet (computer vision + LLM with structured outputs).
- A Streamlit visualisation tool showcases all features.
Tech
Swift (iOS) · Python · OpenAI (LLM, structured outputs) · OpenCV (cv2) · Streamlit · AI video generation (Synthesia) · recommender algorithms
Challenges and learnings
- Navigating the uncanny valley of AI video generation, with few (or no) free APIs.
- Integrating many APIs into one seamless product; planning and scoping as a team.
- The team also launched social channels, positioning FoodieBag as a "food-tok" brand.
Result
The team's Devpost write-up hints at a prize ("we might have bagged a victory"). ⚠️ Confirm which award, if any.
Links
- Devpost: https://devpost.com/software/foodiebag
- Live demo: https://foodiebag.streamlit.app/
- GitHub (team repo): https://github.com/mostafa-elhaiany/FoodieBag
- Linktree: https://linktr.ee/foodiebag