lombarovic

Hi HN, I'm Tomislav. I've been a developer for 15 years, and for about eight I've run Drawize, an online drawing and guessing game I wrote by hand, with hardly any libraries besides jQuery.

This is the story behind Live Coloring Book, which I showed here as a Show HN on 13 October. You color a page in the browser, or print it, color it with real crayons and take one phone photo, and the drawing comes alive in your colors and walks, swims or flies into a shared world with everyone else's drawings.

This time I didn't write the code. I built the first, fully working version in two weeks with AI agents, and I've kept improving it since: Claude in Claude Code wrote the code, and the drawings came from OpenAI's image model through Codex. I was the product owner and reviewed everything by eye. The write-up covers the parts I found most interesting: generating animation rigs from flat line drawings with no animators, reading crayon colors from a phone photo entirely in the browser, and what went wrong along the way.

Feedback I'd love:

sheets the photo reader fails on (bad light, bent paper, old phones); walks or animals that look wrong to you; whether it runs smoothly on your cheapest device. The article was drafted with Claude from the repository and edited by me.

Live now: the mouse wheel zooms the drawing (scroll down again to zoom back out and keep scrolling the page), and there's a "Bigger drawing" button next to the zoom buttons that gives it the whole window. Thanks for the push!

Thanks! Both fair points. Right now zoom is the +/- buttons in the drawing's corner, or Ctrl + scroll (pinch on a trackpad), which clearly isn't discoverable. I'll make the plain mouse wheel zoom the drawing and give it more room on big screens. Will post here when it's live.

Hi HN, I'm Tomislav. I've been running Drawize, a multiplayer draw-and-guess game, for about eight years. This is my new project.

You pick a drawing, color it in the browser, and when you're done it comes alive in your colors and walks (or swims, or flies) into a world with everyone else's drawings. No account, no install required. You can also print a page, color it with crayons and take a photo, and it reads the colors from the paper and makes your paper coloring come alive!

Nothing is animated by hand. Each drawing is split into parts, and a skeleton for its body type (four legs, bird, fish, two legs, and a few more) is fitted to it. The picture becomes a mesh on that skeleton, so legs bend without seams, feet stay planted when walking, and tails and ears bounce with a bit of spring physics. Your colors go into the mesh's texture. It's built with WebGL2, with a Canvas 2D fallback.

The worlds are shared live: when someone sets a drawing free, it walks in on everyone's screen (though you also have private Friends and Family worlds).

Most of it is free; a few worlds and some creatures are unlocked via an optional pass.

I'd really like feedback on the first minute: is it clear what to do, and does the "it comes alive" moment work? I'll be here all afternoon.

Hi HN,

I’m Tomislav, the creator of Drawize (10M+ users ). I’m building Drawize Academy: a virtual drawing tutor that gives feedback on your own drawing while you practice—rather than generating art for you. The first lesson is free to try.

Why I built it

Drawing has been part of human culture since the earliest cave paintings, but today it’s hard to get real feedback: schools cut art budgets and private tutoring is expensive. I wanted to see if software could act as a pedagogical coach that helps you improve through practice.

How the feedback works

Rather than sending only the raw canvas, I augment and preprocess the drawing before it goes to the model. The goal is to give the AI clearer context and make feedback more consistent across attempts, without turning this into an image generator.

The philosophy

This is about "Instructional Scaffolding" — building skill through short, guided practice loops and iteration. We focus on the science of Deliberate Practice rather than "matching pixels" or outsourcing the creative work to AI.

Tech

- Backend: C#/.NET API

- Frontend: React

- Model: Gemini 3.1 Pro Preview

- Logic Layer: Custom drawing preprocessing and augmentation to minimize LLM spatial hallucinations

Early access

I’m currently at the “first sales” wall for this educational model, so I’m offering Early Access Lifetime Access for $19.50 (one-time payment) to validate demand and fund development.

Feedback I’d love

1. Does the real-time feedback loop help you notice mistakes faster?

2. In a world of generative AI, does learning to draw still feel valuable to you?

3. Does the “coach” feel supportive and actionable, or like a grading algorithm?

Happy to answer questions about the .NET/React implementation, how feedback is generated, and the pedagogy behind the lessons!

It is pretty close to that dream scenario now, yes.

Because the tech stack is stable (and fully matured), I almost never have to deal with 'emergency' technical support or bug fixes. The servers just hum along.

I do handle customer support myself, but the volume is very low relative to the traffic. 90% of the tickets are just non-technical questions about billing or ad-free subscriptions.

This low-maintenance overhead is exactly what allows me to work on new features or experiment with new projects (like my upcoming AI drawing school) without burning out.

Thanks! Always good to see other folks sticking to simple architecture — it really pays off in the long run.

Feel free to shoot me an email, though I am currently swamped with the responses here and the Academy launch, so apologies if I am slow to reply!

Good catch! I guess I am still mentally stuck in 2024.

Fixing it now, thanks for letting me know!

Modern servers are absolute beasts if you don't bog them down with serialization overhead and network hops between services.

With efficient code in .NET, a single machine can handle such kind of load without breaking a sweat. I actually sleep better knowing there are fewer moving parts to fail!

Thank you! I am slightly obsessed with optimization, so hearing that means a lot.

You might be surprised — the game is actually deployed in just one region (US) on only two dedicated servers (Contabo).

Here is the breakdown of why it feels fast:

1. The Metal: I use one server for the Web App + Gameplay Backend (.NET), and a second server strictly for PostgreSQL and MongoDB. No virtualization overhead.

2. The Network: I use Cloudflare for static content, which handles the initial global load speed.

3. Aggressive Prefetching: I rely heavily on ServiceWorkers. When you land on the home page, the 'Play' page and game assets are already being prefetched in the background. When you click play, it loads instantly from the local cache.

4. Single WebSocket: Once connected, there is zero HTTP overhead. Every interaction — gameplay, chat, UI updates — travels through a single persistent WebSocket connection.

Keeping the architecture simple (monolith-ish) rather than distributed helps me keep the latency predictable and maintenance low.

Thank you! I really appreciate the kind words regarding performance.

Yes, it is fully self-sustaining. In fact, for the last 5 years, it has been my main full-time source of income, running entirely as a bootstrapped project from Croatia.

The revenue comes primarily from ads, plus a smaller portion from Premium ad-free subscriptions. Since I focus heavily on keeping infrastructure costs low (optimized .NET code + moving storage from S3 to Wasabi), the margins are healthy enough to be a very viable, bootstrapped full-time business.