For anyone out there who ends up creating images for titling (because sIFR is too complex to set up or otherwise) I recommend Cufón (pronounced koo-fone, like “I love the features of the iPhone.” “Yeah, it’s a pretty koo-fone.”)
It creates an image of your text on the fly with javascript. It uses the canvas element in all browsers that support it and for IE it uses VML.
I haven’t run it though any extensive tests but in everything I’ve tried it seems very promising. They are aware of some current limitations, foremost among them being selectable text. They’re working on a solution. Check it out; it may save you some Photoshop time and bandwidth costs.