1) People who can't stick close to time. There's a time limit for a reason - people have other places to go. The last 5 talks I attended all went way, way over time; if you can't get the point across in the time, you don't understand what's important about your work.
Don't these people have ANY idea how long they're going to take? Unless I am interrupted a lot, I can usually time my talk to within a few minutes or so, based off nothing more than the number of slides I have; generally close enough to finish with time for a few questions before the scheduled time is up. I check my watch a couple of times as I go, so if I am running well over for some reason (which almost never happens), I can skip ahead to the more critical stuff. How can someone go almost twice the allocated time without realizing? (And if they do realize and do it anyway, who the hell do they think are?)
2) Morons who think that putting their entire PAPER on slides is acceptable practice. In 12 point.
As soon as I see someone holding their slides in portrait orientation, I start to sweat (well, hardly anyone carries actual physical slides any more - but a few of these guys still do). I'm in for a lot of eyestrain, and they are absolutely guaranteed to run way over time. And they're going to spend the entire talk with their back to me. The back of your head just isn't that interesting.
For some reason, by far the most common offenders on this are LaTeX users. I don't know why. I can read your freaking text 5 times as fast as you can say it. Maybe more. If I want to read your freaking paper, I will read your freaking paper. I do NOT want to listen to you READING your freaking paper to me. This is not going to get me interested in what you did.
If you can't summarize it, you don't understand it well enough to be giving a presentation. If you won't put in the effort to summarize it, why should I put in the effort to sit through it?
3) People who think it's necessary to go through proofs, protocols, algorithms or whatever in excruciating detail. If I want extreme detail, I will read your damn paper. If you did something particularly clever to prove a result, tell me briefly what that was, but otherwise, unless it's the only thing your paper is about* don't do more than present the result, tell me why it's interesting and what you can do with it. Outline the proof if it's really brief.
* and if it is all you can talk about, I probably won't be back at your next talk.
4) If I have to ask "Why should I care about this?" (though I may put it more politely than that) then you have already failed.
That's the entire point of giving a presentation, something you should convey in the first few minutes. If we got to the end of your talk and I am asking, you didn't even hint at it.
If I ask and you still can't answer, you have no business giving a talk on it.
5) People who think that spouting bullshit is better than answering "I don't know". I don't expect every speaker to have a ready answer for every question - that would be ridiculous. "I don't know" is okay, "but I'll find out and let you know" is better. But whatever you do, don't bullshit me. Chances are I will know the moment you open your mouth. If not, I will likely find out a few minutes after the talk has finished. I will NOT be impressed.
6) People who think that brown text on a purple background is a good idea. Black on white is readable. Many other colour combinations are definitely NOT. And fancy wipes and shit - it's distracting, not interesting. Knock it off. Your talk should be interesting. If you have to fuck around with brainbending colours and wipes and fades and crap, you're not spending time on the content. ... and starwipe!
7) People who can't answer basic questions about how they got some particular result (e.g. "Did this come from the survey or the interviews?"). If you don't know your own research that well, why would I believe it's your own work? Even if I do, why would I trust anything else you say about it?
8) People who make some amazing assertion but don't have a good answer for "what is your evidence for this assertion?". Look, if you're going to make a claim, someone is going to make you justify it. If it's an extraordinary claim, you have to have really good evidence.
If you can't justify it, don't make the claim. Just don't.
9) People who put three words and some clipart on a slide and present 5 slides total... (and STILL, unbelievably, somehow manage to go 45 minutes past the allocated 45 minutes).
If you don't have some actual content, you're wasting my time. Have something to say, and don't spend longer than necessary to say it.
10) People who start their talk with a list of assumptions, every one of which is known, beyond a shadow of a doubt, to be completely false in practice, well beyond any sense of "approximation"*. And then finish their talk without attempting to address why their results are in any way relevant, or even identify which assumptions their results are robust to deviations from, and which ones are more critical.
If your work is always a bunch of unexamined, unrealistic assumptions followed by some unenlightening proof of a result and no applications, I am sorry, but I don't think your work is brilliant. I think it's masturbation, with grants.
*("assume a spherical cow" can be just fine, in the right context)
11) People who obliviously break a bunch of these rules, over and over, and never seem to learn a single thing year upon year. Even if you're utterly incapable of self-reflection, at least pay enough attention to notice what works and what doesn't when other people present.