As of now, most of typography-related methods are just prototypes. We added them when we were done with our other Google Summer of Code goals. We do intend to optimize this later.

I don't have much experience with optimizing things like this, but, I managed to trace the source of the slowdown:

Perhaps my observations are obvious to the authors:

1. We can speed up the time it takes np.array() to run by passing it the PIL.Image object rather than PIL.Image.getdata(). Like so:

   data = np.array(self._img, dtype=np.uint8)

   This really only got me from 1.7fps to 14fps on my computer. (640x480 canvas)
2. It might be beneficial if the PIL.Image object's size was constrained to the size of the actual text being printed, rather than the whole canvas size.

Again, I don't really know the fine details of what's going on. I thought I would just point this out.

@MrGallo, you're correct. I didn't look into this much, but your comments do seem reasonable. I don't have the time to work on this right now, but I'll be happy to accept pull requests that implement this change. I'm also assuming that I'm calling PIL.Image.getdata() in other places as well, and that can be fixed too.

It might be beneficial if the PIL.Image object's size was constrained to the size of the actual text being printed, rather than the whole canvas size.

You're correct. Since we were just adding experimental support for text, I didn't spend much time on figuring out the math that gives us a tighter bounding box for the text. Using the whole canvas is really an overkill. Again, happy to accept pull requests :)

This will also be addressed for this year's GSOC (see #104). Might make sense to discuss this there. I'm closing this thread.