Would be more useful on the server. I don't know how it is impplemented, I guess it uses canvas, fortunately there a few serverside canvas implementations available.
I think this library could defer altering the window object until it is certain it is in a browser context. The original design goal was for a browser, but the issue has already come up on leveraging node-canvas to make it useful on the server as well:
"Converts all resources into an Image object by applying the function draw. The draw function receives all resources as canvases..."
I went down the same thought process as you.. Would be useful on the server, coupled with node-canvas.
Wonder whether it would be faster / more scalable than calling out to ImageMagick to add gradients or text to an Image..