JewelMuse

💡 Inspiration

As someone deeply interested in fashion, creativity, and the potential of AI, I wanted to explore how artificial intelligence could empower everyday users to become designers.

Jewelry is a deeply personal and expressive form of art — and I saw an opportunity to make that design process more accessible, imaginative, and interactive using generative AI.

That’s how JewelMuse was born — a tool that turns your aesthetic preferences into custom-designed jewelry pieces, instantly.

🧠 What I Learned

Building JewelMuse helped me learn:

How to design seamless, AI-first user experiences using Bolt
How to craft image prompts for DALL·E to get visually elegant, motif-rich designs
How to structure inputs for creative yet efficient user flows
How to blend generative content with trend-based suggestions in real-time

It also reinforced the creative potential of no-code tools when used thoughtfully with AI.

🛠️ How I Built It

I built JewelMuse on Bolt, a no-code AI app builder.

Key components of the build:

🎨 UI Design: Button-based selection for Type, Category, and Color; dropdown for Design Specifics
🔑 OpenAI API Integration: Users enter their own DALL·E API key (once per session) for secure image generation
✨ Trend Insight Generator: Automatically generates fashion-mag-style insights based on user choices
🎨 Image Generator: Dynamically composes prompts and fetches images from OpenAI DALL·E
💾 Save to Collection: Users can store and revisit their favorite designs within the session — no backend required

⚙️ Challenges I Faced

API Key Handling: Bolt doesn’t have native backend storage, so I had to securely simulate one-time API key entry without hardcoding.
Prompt Engineering: It took iterations to get DALL·E to generate visually consistent jewelry aligned with user selections.
Design Simplicity vs. Functionality: Keeping the UI minimal while offering multiple features (trend generation, image preview, save flow) was a balancing act.
Public Image Embedding: OpenAI-generated images are session-bound, so I had to host images externally to meet submission requirements.