I built an AI product photography tool because I was tired of paying $500 per photoshoot

The Problem

I run a small ecommerce store — about 60 SKUs, home goods and accessories. Last year I spent over $6,000 on product photography. Every new product launch was the same painful cycle: book a photographer ($200-500), wait a week for edited files, realize half the shots don't work for my listings, schedule a reshoot.

Meanwhile the product is sitting in my warehouse not generating revenue because I don't have listing images.

I'm also a developer. And I kept looking at the newer AI image models thinking: this problem is solvable.

What I Built

PixelPanda (https://pixelpanda.ai) — upload a basic product photo (phone snap works), get 200 studio-quality shots in about 30 seconds. The AI handles lighting, shadows, reflections, and background composition.

You pick from scene templates — studio, lifestyle, outdoor, flat lay, closeup, etc. The output is genuinely hard to distinguish from professional photography for most product types.

$5 one-time payment for 200 photos. No subscription.

Why $5 With No Subscription

I looked at every competitor in this space. They all do one of two things:

1. Charge per image ($0.25-1.00 each) — which makes you afraid to experiment
2. Monthly subscription ($20-50/mo) — which charges you whether you use it or not

Both models suck for small sellers who might launch 3 products one month and zero the next. I wanted a price point that's an impulse buy. Nobody agonizes over $5. If you get even 5 usable photos, you saved hundreds compared to a photographer.

The tradeoff is that my revenue per customer is low, which makes paid acquisition hard. More on that below.

Tech Stack

• Backend: FastAPI (Python)
• Image Generation: Flux models via Replicate (Flux Kontext Max for product compositing, Flux 1.1 Pro Ultra for avatar/portrait generation)
• Storage: Cloudflare R2
• Database: MySQL
• Payments: Stripe (one-time checkout, no subscription billing complexity)
• Frontend: Jinja2 templates + vanilla JS (no React/Vue — kept it simple)
• Hosting: Single Ubuntu VPS, Nginx reverse proxy, Supervisor process manager

Total infrastructure cost is about $50/month (VPS + domain). The main variable cost is Replicate API calls — roughly $0.03-0.05 per generation.

What's Working

The product itself works well. I A/B tested AI photos vs professional shots on my own listings:

• CTR within 2-3% of pro photos
• No statistically significant conversion difference
• Users consistently tell me the output quality surprised them

Best product types:

• Hard goods (candles, mugs, electronics)
• Accessories (wallets, bags, jewelry)
• Cosmetics and skincare
• Packaged food products

$5 price point eliminates objections. Nobody asks "is it worth it?" at $5. The conversion from landing page to checkout is decent once someone understands what it does.

What's Not Working

Customer acquisition is my biggest problem. A $5 product means I need very cheap traffic. Here's what I've tried:

• Google Ads: $35/week, ~$0.47 CPC, 1 conversion in 7 days. CPA of $35 for a $5 product. Not sustainable.
• Meta Ads: $40/week, getting clicks but zero attributed conversions from the ad campaign. Just deployed server-side conversion tracking (Meta CAPI) this week so Meta's algorithm can actually see purchases now. Too early to tell if this fixes it.
• SEO/Free tools: Built 26 free image tools (background remover, image resizer, etc.) that get ~177 sessions/day. But only 1.7% visit the signup page. Lots of traffic, very little conversion.
• ChatGPT referrals: This is actually my best channel. ChatGPT sends ~72 views/day with an 11% signup conversion rate. I didn't do anything to make this happen — it just started recommending us for ecommerce product photography queries.

The funnel leaks at pricing. About 28 signups/day, 35% view the pricing page, but only ~4% ever pay. People browse, see the product, and leave. The gap between "this is interesting" and "I'll spend $5" is wider than I expected.

Revenue

Being transparent: I'm still early. About 2-3 purchases per day at $5 each. Monthly revenue is low three figures. Not ramen profitable yet when you factor in ad spend.

The unit economics work if I can find acquisition channels under $3 CAC. Organic/SEO is the long-term play but it's slow.

Lessons Learned

1. Distribution is harder than building. The product took a few months to build. Finding customers has taken longer and is still my main challenge.

2. Low price doesn't mean easy sell. I thought $5 would remove all friction. It removes price objections but you still need trust. People are skeptical that $5 can produce professional-quality photos.

3. Free tools drive traffic but not revenue. My free tools get 5,000+ sessions/week. Almost none convert to paid users. Free users and paying users are fundamentally different audiences.

4. AI-generated content is commoditizing fast. When I started building, AI product photography felt novel. Now there are dozens of tools. The moat is in UX, price, and specific quality tuning — not in "we use AI."

5. Build for the paying user, not the free user. I spent too much time optimizing for signups and free tool usage. Should have focused earlier on what makes someone pull out their credit card.

What's Next

• Trying content marketing and community posts (like this one) since paid ads aren't working at this price point
• Building a Product Hunt launch
• Expanding into video — AI-generated product videos for social media ads
• Considering whether to raise the price or add a higher tier

Try It

If you sell physical products: https://pixelpanda.ai

$5 for 200 photos. Would genuinely love feedback from other founders — both on the product and on acquisition strategies for low-price-point tools. What's worked for you?