The Problem
Real estate agents show empty rooms, architects only have sketches, furniture makers need product photos in various settings. Traditional staging costs €500-2000 per property and takes days. US-based AI tools often lack GDPR compliance and don't understand the European market.
The Solution
Kuvio Studio transforms photos in seconds with specially trained AI prompts: Virtual Staging fills empty rooms, Declutter removes chaos, Twilight mode creates perfect evening atmosphere. One credit = one transformation. Hosted in Germany, GDPR-compliant, no hidden costs.
Results
Architecture
Full-stack TypeScript with TanStack Start for SSR and Server Functions. Google Gemini as primary AI model for image generation. Supabase for auth, storage, and PostgreSQL database – all hosted in Germany. Credit-based billing with transparent pay-per-use structure.
Challenges
- • Prompt engineering for consistent results: Each transformation needs precise prompts that work reliably across different input images
- • Anti-hallucination for furniture: The AI model must not alter existing fixtures – only add or remove what's requested
- • Intensity levels: Three levels (Minimal, Moderate, Full) for each tool, all requiring different prompt strategies
- • DACH market focus: Marketing and UX completely tailored to the German market, not just translated
Learnings
- • Spin-off from maklerupdate: The idea emerged while observing agent listings – empty rooms sell worse
- • Gemini is surprisingly good for image editing: Consistent results, good prompt adherence, affordable API costs
- • Credit system > Subscription for this use case: Users want to pay when they need it, not every month
- • B2B AI in the DACH market: GDPR compliance and German support are real differentiators