I shipped a complete 3D crowd runner in one HTML file. No build, no backend, 68 KB. Live demo, 5 EUR lifetime license, ready to rebrand.
I just shipped something I have been quietly wanting to build for years: a real 3D crowd-runner game that lives in a single HTML file. It plays like the ones you keep seeing in fake mobile ads, but it actually works, on any phone, in any browser, without an app store. It is now live on my marketplace as Crowd Rush v2.0, and you can try the free demo before you spend a coin.
Why one HTML file changes the economics of casual games
Most games in this genre are Unity or Unreal builds targeting the app stores. That means signing keys, review queues, revenue splits, and hosting cost per install. I wanted the opposite: a game that a website owner can drop on their existing host in one afternoon and monetize however they want. So Crowd Rush is built as one index.html. All geometry is procedural, every sound is synthesized live with the WebAudio API, all textures are generated on a canvas at runtime. There are zero external asset files.
The whole download is 68 KB. That is smaller than most JPG hero images. It loads Three.js r128 from three CDN fallbacks (cdnjs, jsdelivr, unpkg) so ad-blockers and private DNS resolvers do not kill the boot, and after that first load the engine is cached forever.
The balance work most fake ad games skip
The genre has a nasty tendency: multipliers spam. Every gate is a x2 or x3 and the crowd explodes into 400 units by level 3, so the game feels great in a fifteen-second ad clip and awful five minutes in. I did the opposite. In Crowd Rush, about 85 percent of gate pairs are additive (plus and minus values) and only around 8 to 15 percent contain a multiplier. When a multiplier does show up, it is always paired with a tempting negative on the other lane, so every gate is a real decision instead of a free win.
The result is that growth feels earned. You are rewarded for reading the field, not just holding the finger down. Boss levels every fifth stage give you a health bar, a stomp attack that kills a handful of units on a timer, and a payout scaled to your survivors.
Ten palettes, one file
Every level cycles through a different color theme. Twilight is the classic pastel purple. Ocean turns the world teal. Sunset paints everything warm orange. Meadow, Cherry, Deep Space, Golden Hour, Aurora, Berry, Emerald. Each palette rebuilds the sky gradient, fog, ground, base, rails, tree tints, obstacle color, and enemy tint. Level intros announce the palette name so players notice the shift. It is a small detail that keeps a very repetitive genre from feeling repetitive.
Persistent progress with no backend
I ship a lot of session-only demos because localStorage feels lazy. In this case it earns its keep. Coins, unlocked levels, all shop upgrades, powers, skins, and per-level star ratings (one to three stars) survive a page reload. There is no signup, no cookie banner, no auth, no server. If a player clears their storage the game resets, and that is fine, because there is no account to lose.
Built to be branded
Every color lives in one PALETTES array at the top of the file. The shop upgrades, skins, and powers are three more arrays a few lines down. Rename the title tag, swap the palette values, tweak the gate balance formula, and you have a completely different game with your name on it. That is why the extended license makes sense: agencies buy one file and ship it to five clients.
What is next
The v2.0 build I put on sale today is production-ready and the demo is public. If enough people pick it up, I will add a v2.1 with two more mechanics I have been sketching: a mid-level checkpoint that saves a partial run, and a daily-reward system that rotates one of the palettes as the "featured" theme. Both fit inside the same single-file architecture. No new engine, no new dependencies.
Try the demo, spend five euros if you like it, drop it on your host and ship the same afternoon. That is the whole pitch.