Inspiration
We live in an era of "lazy" digital communication. A generic "Happy Birthday" text or a pre-made e-card feels cold, efficient, and forgettable. Paradoxically, as AI tools became more common, personal connection started feeling even more robotic.
I asked myselves: Can we use Generative AI not to replace human effort, but to amplify human warmth?
We didn't just want to build another card maker. I wanted to build a digital artisan studio, a place where technology acts as a creative partner (a "Muse") to help people craft keepsakes that feel deeply personal, handmade, and emotionally resonant. KairoPi was born from the desire to merge the nostalgia of a handwritten letter with the bleeding-edge power of multimodal AI.
What it does
- KairoPi is a Personal Storytelling Studio that transforms simple sentiments into breathtaking digital and physical keepsakes. It completely reimagines the greeting card experience through four industry-first features:
The "Living Ink" Engine: It doesn't just use handwriting fonts; it captures the user's actual stroke dynamics in real-time, allowing them to type in their own authentic handwriting with subtle, human imperfections.
Muse (Dual-Core AI): Our Gemini-powered co-pilot that doesn't just generate generic text. It has two modes: "Brainstorm" (for when you're stuck) and "Polish" (an editor that refines your rough draft while aggressively protecting your unique voice and slang).
How we built it
Frontend Core: We prioritized a fast, app-like experience by building a single-page application (SPA) using React (Vite) and Tailwind CSS for our "liquid minimalist" UI.
The Canvas: Instead of heavy WebGL, we engineered a robust DOM-based design stage using react-rnd for direct manipulation (dragging, resizing, rotating). "Living Ink" utilizes dynamic SVG paths to ensure perfect scaling of handwritten vector data at any zoom level.
AI Orchestration (Hybrid Approach):
Text (Muse): Powered by Gemini 2.5 Flash, chosen for its excellent balance of speed and creativity. It adheres to custom system prompts that ensure a high-EQ, non-robotic voice.
Visuals: We use a two-pronged approach. Imagen 4.0 generates high-fidelity artistic backgrounds ("DreamScapes"), while Gemini 2.5 Flash Image handles sticker generation and doodle enhancements.
Video (Kairo Cinema): We integrated the Google Veo API. Crucially, instead of just sending a screenshot, our pipeline constructs a hyper-detailed text prompt describing every element of the user's final card. This allows Veo to creatively "hallucinate" a dynamic cinematic reveal around the described product, leading to far more stunning results.
Challenges we ran into
Balancing AI and Agency: It was tempting to let Gemini write everything. We realized users felt disconnected if the AI did too much work. We had to recalibrate "Muse" to be a hinter and helper, not a replacer, ensuring the user always felt like the primary author.
Accomplishments that we're proud of
It feels magical, not technical. Despite the heavy AI lifting in the background, the user experience feels serene, playful, and analog.
Zero-Shot Handwriting: Successfully creating a system where a user can digitize their handwriting in under 60 seconds without complex hardware—just a mouse or finger.
Hybrid AI Pipeline: Successfully orchestrating three different specialized Google models (Gemini Flash, Imagen 4.0, Veo) into a single, seamless workflow.
What we learned
- Learnt prompt engineering so well, loved working with Ai studio had so much fun
What's next for KairoPi
- Collaborative "Group Hug" Cards: Allowing multiple users to contribute handwriting and doodles to a single office birthday or farewell card in real-time and much more
Log in or sign up for Devpost to join the conversation.