DEV Community

Om Prakash
Om Prakash

Posted on

Solo founder, 3 months in: 85 users, $47 revenue, 71% job success — the honest numbers

#buildinpublic #startup #api #solofounder

Hey IH 👋

I'm Om, building PixelAPI — an AI image processing REST API from my home office in Hyderabad, India. Three months in. Here's the unfiltered state of things.

The Numbers

  • 85 total users (acquired organically, zero paid ads)
  • 9 paying customers
  • $47.64 total revenue
  • 1,085 API jobs processed in the last 30 days
  • 71.6% job success rate (up from 65% two weeks ago)

The Hardware Angle

I'm running this on bare metal GPU servers in my office — no AWS, no cloud GPU bills. That's the whole pricing moat: RTX 4070 + RTX 6000 Ada hardware I already own, running 24/7. Means I can charge 2x below PhotoRoom, Claid.io, and similar services and still have margin.

Operation PixelAPI PhotoRoom
Remove background $0.010/image $0.02/image
Replace background $0.050/image $0.10/image
Remove text/watermark $0.025/image Not offered

Zero cloud GPU bills = structural cost advantage that's very hard for cloud-based competitors to match.

The Failure Rate Problem

28% of jobs were failing silently. Users submitting API calls, getting errors, churning without a word.

Dug into the logs and found three bugs:

  1. Video generation (WAN2.1 14B model) was timing out — the model needs ~70s, we had a 30s hard limit
  2. remove-text endpoint had a Python variable shadowing bug — io was imported as a module then overwritten by a local variable called io
  3. remove-bg was hitting CUDA OOM on large images with no graceful fallback

Fixed all three. Success rate now trending toward 90%+. Monitoring daily.

The lesson: silent failures are the worst kind. A user who hits an error and gets no explanation doesn't file a bug report — they just leave.

What I'm Not Building (The Anti-Roadmap)

Inspired by the Photoroom anti-roadmap post in today's Indie Hackers newsletter, I wrote my own.

PixelAPI will NOT build:

  • ❌ A visual editor or no-code tool (that's Canva)
  • ❌ A stock image library
  • ❌ Image generation (50 APIs already do this)
  • ❌ Enterprise sales / custom contracts
  • ❌ Mobile SDKs (yet — after $10k MRR)

The constraint forces clarity: we're a developer REST API, nothing else. Every feature gets filtered through: does this make integration faster, cheaper, or less friction? No → don't build it.

What's Next

  • Get job success rate to 95%+
  • 10 paying customers
  • $100 MRR
  • Better SDK examples + docs

If you're building something that needs image processing — background removal, watermark removal, upscaling, face restoration, video generation — try it out: pixelapi.dev — 100 free credits on signup, no credit card needed.

Building in public from Hyderabad 🇮🇳. Drop a comment if you're doing the same.

— Om

Top comments (0)