I have no authority to give you any advice on this issue, although there are some general tips to keep in mind. (a very nice discussion on this topic can be found here, or, you know, google it)
The first thing to consider is the kind of audience you have to face. I'm personally not a fan of changing your scientific message too much for different audiences, but academia may require different levels of formality. The best tip on this matter, in my opinion, is to let your personality shine in any situation, a talk from someone feeling comfortable is much more enjoyable.
Nobody, in principle, cares about your research as much as you do. Therefore, your first goal is to make the audience care, not to show them how good you are. First, if it's not clear why they should care about your obscure research, they will never fully appreciate the beauty of your methods or the elegance of your results. Secondly, and more importantly, a talk is not a lecture, you are not there to teach them how to do your job. It seems rather trivial, but it is not uncommon to find speakers that, being too focused on showing the details, completely miss the main message of their talks. It is also obvious that 99% of the audience will not probably understand your over-detailed presentation, although nobody usually admits that, scared to look stupid. If you want to give a great talk, you should attract the audience to your topic, drive them through the most basic steps till they care about your problems, show them your ideas, not your tedious technical steps. In this way you will achieve the real goal: make them feel smart as much as you (and more than you, they figure it out in less time than you).
All of this should be obtained by showing also your humble smartness. A piece of cake!
Assuming that you succeed in this extremely difficult task (especially the first 26 times), here come the questions from the audience. You can measure the effectiveness of your talk from this moment. A good strategy is to design your speech in order to drive the attention to specific questions of your choice. In this way, you will look smart and prepared on your research, the audience will feel more involved in your scientific process and you will take out some time from the possible nasty question waiting for you in some dark alley.
To fully accomplish that beautiful scenario, you need to know some of the people in front of you. Maybe there is an expert in a different field, known to always ask something about his/her research at any given seminar. As I said, do not change your scientific message, but be prepared to that predictable question, it will show your interest to topics other than yours.
Sadly, not all the questions will be reasonable or come from an innocent voice. The tip that never failed so far is: never engage.
There can be the one that wants to show off and not really ask questions, but rather talk for a minute about something. This person will probably look smart to everyone except to you, because if you are the speaker you can (and you have to, in order to not being intimidated) spot an idiot. But the rule is: never engage, just thank for the comment and move on.
Occasionally, you may find one of those people that translate their smartness in being a flaming asshole. Those that will begin their question with a very polite I don't agree with everything you said or ask you to step way outside your comfort zone just to catch you off guard. Being an asshole to look smarter is sadly a common habit in academia, but you are even smarter and you remember the golden rule
If caught off guard, do not panic, discuss it during the coffee breakThe reason is that, let's say 95% of the times, you are the only one with a microphone, the only one that everyone can hear (if so, remember also to repeat the questions, it gives you time to think and you let everyone participate). Moreover, flaming assholes are not there to discuss with you, they just want to climb the ladder by stepping on you. Therefore, if you engage a discussion with someone, you will most certainly look like a crazy person speaking to the wind about something that no one else could possibly be able to understand.
Considering that there is a possibility that you looked exactly like that during your talk, it is better to give a different impression during the question time.
You always have to remember that you are there to speak to everyone, use your power (the microphone) to make that happen. You will give a better impression, you will keep full control of the situation and, if you don't know the answers, nobody will really hunt you down during the coffee break: it is the bloody coffee break, the only religion tolerated in academia.