What you're trying to do sounds very like the process used to generate CAPTCHAS, and that's something that can definitely be done by web services. Whether it's the most convenient way for you to do it is a more complicated question.
Certainly turning text into an image shouldn't need OpenGL, which is a general 2D/3D graphics library and is aimed more at real-time 3D graphics (e.g. games) than generating static images (I'm not saying it can't be used for static images, just that's not really its focus). I would suggest looking at PIL, the Python Imaging Library, as I'm pretty sure it has primitives for rendering text as a static image. It also has the advantage of being pre-installed on the PA environment, I believe.
What I'm not sure about is how easy it'll be to get an image from PIL into whatever format numpy.ndimage requires - I'm sure it must be possible to get it into plain bitmap data form, however.
Edit: Whether PA is the best place for you to do this stuff is really a question for you to answer yourself. As I say, it has pre-installed tools for dealing with images, but if they're not the ones with which you're familiar then you may burn quite a bit of time learning them. On the other hand, if you've got the time then it might be a useful way to expand your Python knowledge, if that's something that attracts you. Certainly PA makes it very convenient to host web applications which might be useful to you. If you need to do any real-time (i.e. animated) graphics then I'd strongly consider instead running scripts locally instead. Nothing to stop you using PA as a shared editing environment even then, especially as it makes integrating with things like GitHub easy (assuming you want source control, and I'd really definitely recommend it for even small projects).