Hey everyone,
I'm a full-stack developer and have been fascinated by the progress in LLMs and image generation for the last year and a half. After seeing the AI headshot market explode, I noticed a common complaint: you'd pay for 100 images just to get a handful of good ones.
I knew I could do better by focusing on model quality and prompting. So, I challenged myself to build and launch a high-quality alternative from scratch in just two weeks.
Today, I’m excited to share that I launched Smart Headshots AI.
My 2-Week Sprint:
- Frontend: Next.js (love the performance)
- Backend: Supabase for everything - auth, database, storage
- UI: Kept it dead simple internally, went all-out on the landing page using shadcn UI + Tailwind CSS
- Payments: Switched to Dodo Payments (MOR) after hearing too many Stripe horror stories
- Architecture: Everything runs through webhooks - payment providers talk to AI APIs seamlessly, deployed on vercel.
I use Cursor, so first I integrated supabase and dodo payments MCP servers, and created all required functions for the payments, auth and database. I spent more time on creating landing page since it's the first door that your customer will see. Analyzed few competitors and finalized one wireframe, provided it to claude sonnet 4.0 and voila! It generated very beautiful landing page, of course I spent few hours optimizing the landing page. Spent few more hours to make everything mobile responsive. I used Replicate for AI APIs.
What I Focused On:
- Quality over Quantity: My main focus was on fine-tuning the model and prompt engineering to ensure a very high hit rate of professional, usable headshots.
- User Experience: I kept the internal dashboard dead simple. Upload photos, pick your styles, and that's it. All the complexity is hidden. I poured most of my design energy into the landing page to build trust and clearly explain the value.
- Landing Page: First impressions matter - made it convert
The Technical Stuff That Worked:
- Supabase is great for rapid development
- Webhook architecture kept everything decoupled and scalable
- shadcn components saved me probably 20+ hours of custom CSS
- Dodo Payments integration was surprisingly smooth
What I Learned:
- Ship fast, iterate faster
- Don't over-engineer v1 (guilty as charged on previous projects)
- Payment processor choice matters more than you think
- Quality beats quantity every time
For those who missed the link here it is: https://www.smartheadshots.ai/
Thanks for reading this far! I'm happy to answer any technical questions about the build process. If anyone would like to try let me know I can share few discount codes😉
Top comments (0)