DrawLa: draw while an AI guesses along
DrawLa is a free drawing game for the browser. You get a word, you get 30 seconds, and a neural network tries to recognise live what is appearing under your pen. The sooner it gets it right, the more points you score. No account, no installation: pick a nickname and start drawing.
How a round works
At the start of a round your word appears — something like "bicycle", "cat" or "lighthouse". Only you can see it. As soon as you put down the first stroke, the recogniser re-evaluates your drawing several times per second and shows its current guess along with how confident it is. So you can see in real time whether you are on the right track, or whether your bicycle is stubbornly being taken for a pair of glasses.
Points depend on when the AI hits your word. Whoever manages with a few well-chosen strokes comes out ahead — not whoever draws most neatly. After every round you see all the pictures side by side with their recognition scores, and vote on which one was best.
Three game modes
Solo
Start immediately against two bots. No waiting for other players.
Public
Matched with others automatically. Once four players are in, the round starts by itself.
Private
Open a room, share the code, play with friends. The code works on phones too.
Chaos effects
Optionally, the game interferes. From "Mild" to "Insane" you set how often an effect strikes: confetti in front of the canvas, ants stealing your ink, mirrored controls, strokes that suddenly wobble. If you would rather draw in peace, switch the effects off entirely — they are an ingredient, not an obligation.
What the recognition can and cannot do
Behind DrawLa sits a network trained on Google's Quick, Draw! dataset — millions of real scribbles from around the world, spread across 345 words. It therefore recognises exactly what people usually draw when they have to be quick: stick-figure logic rather than anatomy.
That also explains the typical failures. Draw with particular care and you are often recognised worse, because detailed pictures are rare in the training material. Unusual perspectives barely register, and related shapes get confused: snail and seashell, cloud and sheep. How the model is built and why we use a CNN rather than an RNN is described in the making-of.
Common questions
Is DrawLa free?
Yes, completely. There is no paid content and there are no subscriptions.
Do I need an account?
No. You pick a nickname, that is all. There is no sign-up and no password.
Are my drawings stored?
No. They live in the server's memory only while the game is running and disappear afterwards. Details in the privacy policy.
Does it work on a phone?
Yes, in the mobile browser. Drawing with a finger works well — many rounds are won with rough shapes anyway.
How many people can play?
Up to four per room. If players are missing, bots fill in so that a round never fails for lack of company.
Why doesn't the AI recognise my picture?
Usually because it is too detailed. Reduce it to the silhouette and the one or two features that identify the word instantly. More tricks in the game help.