A high-performance AI pantry chef web app powered by Gemini 2.5 Flash. Synthesizes procedural cooking fry/sizzle soundscapes, runs interactive Cook Mode with voice guidance, and ranks custom recipes based on safety, staples, and dietary constraints.
- Client
- FlavrSelf-initiated project
- Year
- 2024 - 2026Live on Vercel
- Role
- Lead Full-Stack DeveloperFrontend & AI Integration
- Stack
- React 19 · Vite · Web AudioGemini API · Vercel
- Status
- LivePublicly accessible
Flavr was born from the common kitchen dilemma: having miscellaneous ingredients but no clear recipe. Standard search engines suggest recipes requiring missing ingredients, leading to grocery runs.
By leveraging the Gemini 2.5 Flash model with strict JSON schema outputs, Flavr performs instant safety checks (preventing toxic combinations) and generates beginner-friendly recipes with dynamic substitution recommendations.
The UI is built with React 19, Vite, and Tailwind CSS. It features an interactive Cook Mode with step-by-step voice guidance (TTS), built-in cooking timers with mechanical tick sounds, and a custom procedural Web Audio synthesis engine that generates realistic frying and oil bubbling soundscapes while the chef is cooking.
Pantry to plate,
optimized by intelligence.
To build a zero-waste cooking companion with zero latency. No signup barriers, no advertisements—just add what you have, quick-add common staples, and let AI design your culinary journey.
Every generated recipe includes approximate nutritional highlights (calories, protein), ingredient matching analytics, and step-by-step cooking telemetry.
Streamlined UI,
uncompromising simplicity.
Users tag ingredients with autocomplete and redundancy warnings to avoid duplicates (e.g. 'onion' vs 'onions'). The system filters by dietary restriction (Vegan, Vegetarian, Keto) and meal type (Breakfast, Lunch, Dinner).
A step-by-step Cook Mode provides active visual/timing cues, mechanical timers, and speech synthesis (TTS) reading steps out loud, creating a hands-free kitchen assistant.
Interactive 3 cooking tools.
React frontend
Gemini orchestration.
A client-side React 19 application orchestrating Gemini structured JSON responses and synthesizing real-time kitchen audio landscapes via the Web Audio API.
- React 19
- Vite
- Tailwind CSS
- Web Audio API
Modular UI component architecture managing pantry lists, favorites (persisted locally), active cooking steps, and procedurally generated sizzling sound loops.
- @google/generative-ai
- Structured JSON Schema
Configures responseMimeType to application/json on Gemini 2.5 Flash to guarantee type-safe recipe models containing nutritional highlights and instructions.
Cooking made
accessible, waste reduced.
Engineered by byteWizard-zero.
- Creative Technologist
- Zenith Soumya
- CSE Student @ ITER
Zenith Soumya is a Computer Science & Engineering student at ITER College who designs optimized DSA engines, IoT architectures, and agentic AI systems.
Flavr is deployed publicly on Vercel, demonstrating how client-side Web Audio API and Gemini structured outputs can create immersive, zero-overhead utility applications.