Hey CPW1,
I think in terms of speed mixed with comprehension, you're much worse off reading out loud. The subvocal barrier being somewhere between 600-800, though most people can't subvocalize past 200-300 wpm. As you have heard, Photoreading builds speeds up to 25,000 wpm and increases comprehension.
When you image stream, whether you do it outloud or not, you're going to be speaking (or subvocalizing) at the same speed, so you have no choice. The reason you do it outloud is to describe it to some outward force in order to not worry about remembering what you're saying, so that you can "decode" it, and as a result of that, your image stream will keep...hmmmmm.... "streaming."
-JackTuff13