Foosion AI

• 

  Home Page

What inspired me

From the moment I tasted mismatched yet harmonious flavors in fusion cuisine, I was fascinated by how surprising ingredient pairings could create culinary magic. I wanted to democratize that sense of discovery. Combining AI—recipe generation, step-by-step slide‑videos, plus an AI voice assistant—felt like the perfect mix. It made sense to build this directly in bolt.new for rapid prototyping and seamless interactive features.

What I learned

AI-Powered creativity: I came to appreciate how generative AI can invent flavor combinations I’d never consider, while still needing human curation for coherence.

Interactive storytelling: Structuring the user flow—slide‑by‑slide guidance synced with AI voice—taught me how timing and pacing deeply affect comprehension and enjoyment.

UX in cooking contexts: Kitchen users need hands‑free control, clear visuals, and minimal friction; building touch‑and‑voice‑enabled navigation was a game‑changer.

How I built my project

Core engine: Hooked into a generative‑AI API to produce fusion recipes based on user inputs (cuisines, ingredients, dietary needs) powered by Dall-E, Gemini AI and ElevenLabs AI voiceover.

Bolt.new setup: Created a multi‑stage slide deck where each step (e.g., “chop garlic,” “saute onions”) is represented visually and narratively.

AI voice assistant: Leveraged bolt.new’s audio block capability and TTS API to narrate instructions. Users can pause, skip back, or ask for clarifications.

Interactive features: Added dynamic timers, ingredient checkboxes, and “explain flavor hint” overlays to deepen user engagement.

Challenges I faced

API hallucination: AI sometimes invented nonexistent steps or mismatched ingredient quantities. I mitigated this with validation logic and fallback prompts.

Syncing audio and visual flow: Ensuring the voice aligned precisely with slide transitions required manual tweaks and introducing delay buffers.

Voice UX balance: Users didn’t want to hear narration for trivial actions, so I added sensitivity thresholds to skip non-essential prompts.

Mobile responsiveness: Bolt.new’s layout didn’t always render well on small screens. I used responsive containers, collapsible sections, and thumb-friendly controls.

Strange behavior: When I tried to provide a prompt such as numbered list, sometimes bolt.new AI collapsed the entire site and started to rewrite from scratch, fortunately I had a backup repo in GitHub.

Overall, this project has been a whirlwind of culinary creativity and technical problem‑solving. It blends the joy of cooking with the power of AI, all wrapped in an accessible, interactive experience built entirely in bolt.new.

Built With

• ai
• api
• typescript