Drinkable

🍺 About the Project

Inspiration

The idea for Drinkable came from a simple observation:
Most people don't actually know how fast they’re getting drunk.
Existing BAC calculators only show numbers, not insight. They don’t reflect how you feel, how quickly your BAC is rising, or whether you’re entering a risky zone.

Our team wanted to build something that feels friendly, intuitive, and personal—a tool that makes alcohol safety approachable through visuals, not just data.
This inspired us to design Drinkable: an AI-powered web app that transforms your BAC into an interactive character that reacts to your drinking pace.

What We Learned

Throughout the project, we explored:

How real-time BAC is calculated using the Widmark formula
[ BAC = \frac{5.14 \times A}{W \times r} - 0.015 \times H ]
How alcohol absorption behaves over time and why drinking speed matters more than total quantity
How to build a dynamic UI that updates in real time as BAC changes
How AI vision models can detect drink types and estimate alcohol content
How to create a consistent design system across characters and icons
The importance of UX clarity when communicating safety recommendations

We gained experience in frontend architecture, backend APIs, and AI-driven interactions—all under strict time pressure.

How We Built It

Our stack and architecture:

Frontend

React + custom UI components
Real-time character rendering tied to BAC states
Flat + soft-3D icon system (beer, soju, whiskey, wine, cocktail, water, heart, speech bubble)

Backend

Python + FastAPI
BAC computation engine
Timestamp-based drinking pace analysis
Safety recommendation generator (e.g., “Wait 12 minutes before your next drink”)

Natural language generation for safety coaching

Database

User profiles
Drink history
Timestamps for each drink
BAC curve logs

Everything works together to create a responsive drinking assistant that updates every time the user adds a drink.

Challenges We Faced

Building Drinkable in a short hackathon window brought several challenges:

1. Designing 10+ Character States

We needed a character that could express intoxication clearly without judgment.
Balancing friendliness, clarity, and consistency across 10 stages was difficult.

2. Real-Time BAC + Timing Logic

Implementing BAC rise and fall over time required tuning:

absorption period
elimination rate
timing for recommendations
safety thresholds

Ensuring the math felt natural in real use was a major challenge.

3. AI Vision Reliability

Detecting drinks from messy photos was harder than expected.
Lighting, angles, and reflections all affected predictions.

4. UI Consistency

We created a full set of alcohol icons from scratch.
Matching tone, color, and perspective across all icons was time-consuming.

5. Scope Management

With so many ideas—food logging, social mode, hydration tracking—we had to carefully choose what was feasible within the hackathon.

Conclusion

Drinkable taught us how to combine math, AI, user experience, and design into one cohesive safety tool.
We hope it helps people understand their drinking habits better—not through judgment, but through a friendly character and accessible insights.