Frequently Asked Questions

Does this use AI or machine learning?

Not in the modern sense. Genetic Painter uses a classical evolutionary algorithm — there are no neural networks, no training data, and no pre-learned models. The algorithm works purely by trial and error: randomly painting shapes and keeping only those that improve the result. Everything runs locally in your browser using the HTML5 Canvas API.

How long does it take to recreate an image?

It depends on the image complexity and your shape settings. Simple images with large color blocks may look recognizable after a few thousand generations. Detailed photos with many colors and fine textures may need 50,000 or more accepted shapes to achieve a faithful recreation. You can stop and admire the result at any point — many users find the "in-progress" paintings to be the most aesthetically interesting.

What kinds of images work best?

Images with strong contrast and distinct shapes tend to produce the most striking results quickly. Portraits, landscapes, and images with bold colors work particularly well. Very detailed or low-contrast images will eventually be reconstructed but may take longer to become recognizable.

Is my image uploaded to a server?

Your original image never leaves your device. All processing happens locally in your browser using JavaScript and the Canvas API. The app does periodically save your painting's progress to a server so you can resume later, but the original source image stays private on your machine.

Can I save my painting?

Yes — right-click (or long-press on mobile) the "Painting" canvas and choose "Save image as" to download your current result as a PNG file. Your progress is also automatically saved to the server every 500 accepted generations, so you can close the tab and return later to continue where you left off.

Why does the acceptance rate slow down over time?

In the beginning, the canvas is blank, so almost any colored shape will reduce the overall difference from the target. As the painting becomes more accurate, there are fewer "easy wins" — the algorithm needs to find shapes with just the right position, size, color, and transparency to make a measurable improvement. This diminishing return is a natural property of optimization algorithms converging toward a solution.

Why does the painting look blurry at first?

When using large radius values, the algorithm paints broad areas of average color first — this is the most efficient way to reduce the overall difference from the original image. Think of it like a real painter who blocks in large shapes of color before adding detail. Reducing the max radius forces the algorithm to use smaller shapes, which capture finer detail.

What image formats are supported?

Genetic Painter accepts any image format your browser supports, including JPEG, PNG, GIF, WebP, and BMP. For best results, use images under 1000 pixels wide — larger images work fine but will run slower since more pixels need to be compared each iteration.

Is this the same as AI image generation?

No. AI image generators (like diffusion models or GANs) learn patterns from millions of training images and generate new images from text prompts. Genetic Painter is fundamentally different — it uses a simple optimization algorithm to reconstruct a specific image you provide. There is no neural network, no training data, and no AI in the modern sense. It's pure mathematics: random mutation plus selection.

Can I use this for commercial purposes?

Genetic Painter is free to use. The paintings you create from your own images are yours. However, please ensure you have the rights to any source images you use.