I Ran My Released Game Through My AI Game Cost Calculator: Prediction vs Reality

Jonhisking · October 8, 2026

I'm a solo developer from Korea with zero coding background. I make games by describing them to an AI. My first one, Sperm Race (a silly sperm racing game), is live on Google Play. People kept asking me "how much does it cost to make a game with AI?", so I built a free calculator.

But I didn't actually know if my own calculator was right. So I fed it the exact specs of the game I had already shipped.

Sperm Race screenshots

What I entered

Genre: racing · Size: small · Art: simple vector · 2D
Content: 3 characters, 3 music tracks, 13 languages
Features: ads, in-app purchase, save data · Platform: Android · Engine: Phaser (in reality, plain HTML5)
Tools: Claude Pro, Pixabay (free music), free sound effect libraries

It's all buttons. The only thing I typed was the game's name.

The calculator's input form

Prediction vs reality

	Calculator	Actual
Cost	$47–102	$45 (1 month of Claude Pro $20 + Google Play fee $25)
Time	19–31 days	~17 days (3 days to build + 14 days of bug fixing)
Code	~5,400 lines	~3,100 lines at launch → 13,800 now (after adding multiplayer)
Art	$10 (Midjourney)	$0 (the AI drew characters and backgrounds in code)
Music & SFX	$0	$0 (music from Pixabay, sound effects generated in code)

The calculator's result for Sperm Race

What it got right

The cost was right almost to the dollar. The only gap was the $10 for art, and that's because I had the AI draw everything in code instead of generating images.

What it got wrong

Time: the calculator was too pessimistic. The game itself took 3 days. The other 14 days were almost all bug fixing during Google Play's closed test. When I asked the AI to find a bug, it often "fixed" something unrelated, so I ended up hunting most bugs down myself. (The 14 bugs, one by one)
Lines of code: too high for the first version. But counting updates after launch, I've blown way past it.
"This plan's usage limit is likely too low" warning: in my case, Claude Pro was enough to get to launch.

How it calculates (roughly)

Size is anchored to real games built solo with AI coding agents: small about 2 weeks, medium about 4, large about 6. Genre, features and number of languages move it up or down.
Images = characters × animation frames + backgrounds + items + UI. It assumes you keep 1 of every 3 images you generate, so it plans for 3× as many generations.
Coding tokens assume about 20 million tokens read per working day (mostly cached context) for 70% of the schedule. The 343M here is that estimate. I was on a subscription, so I never measured my real token count.

What I learned

With AI subscriptions, money goes out by the month, not by the token. Every month you slip is another bill.
Plan more time for fixing than for building. For me it was more than four times as long.

The calculator is free, no sign-up. Pick a genre and size and you get the list of images and sounds you'll need, a cost and time estimate, and a ready-to-use starter prompt. It also handles 3D and Roblox games now.

If you've shipped something, try your own project and see how far off it is. I'd love to know.

Pick a genre, size, engine and platform to get your own cost, schedule and ready-to-use prompts.

Open the AI game cost calculator →