Why I Built This
Most AI tools force subscriptions. I wanted pay-per-use.
I use AI image tools constantly for client work and personal projects. Every tool I tried had the same problem: you pay every month whether you use it or not. Midjourney, Adobe Firefly, DALL-E all subscription-based, all wasteful for designers who generate in bursts.
So I decided to build it myself using AI coding tools to handle the technical side, while I focused on the product decisions, UX, and architecture. This is how designers build software in 2025.
Every AI image tool required a subscription
Best models were fragmented across platforms
Pricing was never transparent
This was also a personal challenge to go from Figma to a live, revenue-generating product without a technical Knowledge, using only AI coding tools.
Tools I chose and why
each one earned its place.
Every tool was chosen for a specific reason. Nothing was added because it sounded impressive. Everything had to earn its place in a solo-built product.
AI Models Integrated
How I Built It
Six stages of building a
real product as a designer.
The process wasn't linear I designed, broke things, debugged, redesigned, and shipped. Here's the honest version of how this went.
1 – Product Architecture
Defined the full system before touching any tool. Credit model, FIFO expiry logic, multi-provider routing, Supabase schema, API endpoint structure. Product spec first.
2 – UI Design in v0
Designed every screen in v0 auth, dashboard, generate page, gallery, pricing, sidebar. Used ImagineArt + Freepik as visual references. Dark theme throughout.
3 – Backend via Supabase
Full-feature logic credit deduction (FIFO), Supabase triggers for free credits on signup, Razorpay order creation and webhook verification, API route structure.
4 – Model Integration
Connected 10+ models across fal.ai, ByteDance BytePlus, and Google Gemini API. Each provider required its own request schema the hardest part of the technical build.
5 – Debug + Deploy
Used V0 to run and debug. Fixed build errors, font import duplicates, JSX syntax issues, and provider payload mismatches. Every error read and traced manually.
6 – Live + Framer Site
App deployed on Vercel. Landing page built in Framer. Razorpay business onboarding completed with all compliance pages. App is live and accepting real payments.
Features Shipped
Everything a SaaS needs. Built solo, no shortcuts.
These are not wireframes or prototypes. Every feature listed here is live in production and has been tested with real usage.
10+ models, one interface
Generation
Model selector grouped by provider. Each model has its own credit cost, aspect ratio support, quality options, and API schema. Text-to-image and image-to-image modes. Batch generation up to 4 images. Credit deduction happens before generation refunded automatically on failure.
FIFO credit system with expiry
Credits
Each credit purchase is a separate entity with its own expiry date. Credits are consumed oldest-first (FIFO). Starter pack expires in 1 month, Power pack in 6. Total balance is calculated dynamically never stored as a single number.
Full business compliance stack
Business
Razorpay merchant account setup, purpose code P0803 (data processing), SaaS business category selected, Terms, Privacy, Refund, Shipping, and Contact pages written and published. Google Cloud credits applied for model inference cost offset during development.
What Got Built
What Broke
Every provider had a different rule. Every rule had to be learned the hard way.
Each solution addresses a specific constraint from research. The goal across all of them was the same: one system that operators could use without thinking about the system.
