Built an AI product photo tool in 30 days because the existing ones warped labels. Launching today.
TL;DR — Pikstage (https://pikstage.com) is an AI product photography tool with a real moat: brand-locked consistency. We preserve your product's identity (label text, colors, shape) across 25+ generated scenes. Competitors break past ~10. Built in 30 days, $0 raised, solo founder in Vancouver.
The honest story:
I tried Pebblely, Flair, Photoroom, and Booth AI on the same candle photo. All four warped the label text by photo 8. For a Shopify seller, that's a non-starter — your product looks "mostly right" which is worse than no AI photo at all.
The technical fix is IP-Adapter image conditioning (vs. pure text prompting). I spent Day 4 of the build testing 4 different models with Claude scoring identity preservation. Flux Kontext Pro scored 8.25/10. Locked it in.
The stack:
- Next.js 16 on Hetzner ($5/mo)
- Supabase for auth + DB + storage (free tier)
- Replicate for image gen (~$0.05/image — this is where the costs go)
- Stripe + Resend + Cloudflare
- Total month-1 burn: ~$60 + image API costs
Pricing:
- Lite: $9.99/mo / 25 photos
- Starter: $14.99/mo / 100 photos
- Pro: $39/mo / 500 photos
- Agency: $149/mo / 2,000 photos
Unit economics: ~$0.05 cost per photo, charging $0.40/photo at Lite, $0.15 at Starter, $0.078 at Pro. 60-85% gross margin from day one.
What's hard:
- Distribution. I have a working product and zero customers. That's why I'm here.
- Onboarding. Customers don't know what "IP-Adapter" means or why to care. Wedge is hard to explain in <10 words.
- Glass / reflective products are still wonky. Working on it.
What I'd love from this community:
- Brutal feedback on the landing page copy
- "Here's the SKU you should test on" — happy to run free generations for any IH member
- DMs from anyone who's solved the "explain a technical moat in 5 seconds" problem
7-day refund, 1-click cancel, no card to start. pikstage.com
Building publicly here. Happy to share month-by-month numbers if anyone's curious.
— Dev
