I think, bizarrely, the control pattern for mute "flows" the opposite way to the control for pause/play. If there's a play icon, and you click it, you expect it to play the video. If there's a pause button, you expect it to pause the video. Going against this pattern, if I see a speaker icon, and I click it, I expect it to transition to a crossed-speaker, and mute the video, and vise versa. That is: the current button state in play/pause reflects the action of the button, while the current button state in volume reflects the state of the button.

I think these conventions are easily solved for video, anyway, because it's obvious when the video is muted or not (hopefully).