Oh, I've assumed that you're using a game system that lets the players make a lot of choices about what kind of characters they want, like Steve Jackson Games' GURPS. If you're using something more strict, such as TSR's AD&D game, much of what I say here won't be applicable, since the rules don't give you any choice. Personally, I use the General Goodness Points system for my campaigns, but I suspect that that system is only really usable by experienced gamemasters.
Take another long, hard look at them. If the campaign works out, you and your players will spend a lot of time with these characters. The first Amber campaign I ran had taken somewhere between 300 and 400 hours of game time spread over nearly two and a half years when it ended. That was not an unusually long campaign. Some players made quite good characters from the start, and they kept those characters right through to the dramatic end. Some players didn't make as suitable characters, and they dropped out or switched characters after a while since they didn't have any (or at least not as much) fun. And after all, having fun is what it is in the end all about. Take care when making characters. Think about them. Let it take time. Try to figure out if they're suitable for the campaign you've visualized. Decide if you really want them loose in you world.
If (or, rather, when) you find something you don't like, first of all ask the player if you've correctly understood what they meant. Sometimes it turns out that what they actually wrote is far from what they thought they had written, and their actual thoughts were just fine. Sometimes they can't explain what they were thinking. In that case, they probably won't have much against changing whatever it was you didn't like.
Be nice to your players while you're finding faults in their characters, and be prepared to give in to them if they feel strongly about something. Hard feelings aren't fun, and my experience is that if the player feels strongly enough about some aspect of a character it will work out no matter what.
Try to argue your players out of their worst follies as far as it's possible, but be prepared to say "No, you can't have that!" sometimes. For example, in some version of the GURPS rules, it was possible to design a starting character who was able to juggle houses telekinetically. No GM with their sanity intact would allow such a character into their campaign (unless, of course, it was a suitably deranged campaign).
In the end, there should be a set of characters with whom both you and your players are comfortable. Most probably, it's also time for another pot of tea.
It's time for intros. It's rather like the first draft of a text; you think everything is finished, but as soon as you look at it for real you realise that there are still things which need fixing.
Get everything ready as if you were starting to play. If things work out, you will start playing. Often it's hard to tell when the intros end and the "real" game begins.
For me, getting ready mostly consists of making sure there's tea, that I have the right notes in front of me and getting the players to stop discussing Aristotle (or whatever). When all is ready and you have your players' attention, you can begin.
As I hope you remember, we said earlier that we would start in a
tavern, and that we'd ask all players to describe how their character
came to be there. Now is the time to do that asking. Chose a player
and ask, using your best "I'm the GM, thus God" voice:
"Why are you at the tavern, and how did you get there?"
Don't expect a very good answer unless the player in question is a
hardened veteran roleplayer.
You've just given that player the honour of starting the campaign. She will, to a fairly large extent, set an example for the other players on how to answer your question. Try to force her to give a good answer, preferably in such a way that she doesn't notice.
That is far easier said than done.
The best way I have found to get a good answer is to ask questions. Lots of questions. Listen carefully to what she says and ask questions about details. Ask her to elaborate. Ask about background stuff, about why things are the way she describes. Keep the questions coming until you think it's starting to get ever-so-slightly dull, or the player seems to start getting uncomfortable. Then wrap up, to bring the character's story all the way to the tavern (if it hadn't already gotten there). Then go to the next player and repeat the same thing again.
The concept of "firmness" is, I think, vital to gamemastering, so I'll digress a bit here.
Some things in your gameworld are firm, and some are not. Firm things are those that the PCs have actually axperienced themselves, things they have seen with their own eyes. Such things can't be changed without a valid in-world reason. If a room had one window the last time the PCs were in it, it'd better have one window the next time they enter it too. If you want it to have another, or none, there must be some explanation of how the new window got there (or how the old one vansihed, as the case may be). If, on the other hand, the PCs had only heard someone say that there were one window in the room, you could change your mind about the actual number of windows at any time until they actually saw the room, without an explanation for a change. In the latter case, the number of windows weren't firm.
So how does this matter, you ask? Well, it means that you can adapt the world to the players on the fly. Imagine that they are walking along a road, on their way to a deserted castle, and that they come to a crossroad. They can proceed along three different ways. Or, at least, so the believe they can.
If the crossroad is firm, if the players have walked that way before, they can indeed choose which way to proceed. But if it's the first time they're there, if it isn't firm, they can only come to one place no matter how they chose: to wherever you want them to wind up. If they go left, that's the way to the deserted castle. If they go right, that's the way to the deserted castle.
Do you understand?
So, this first time, don't be too concerned with the plot. Give the players time to become familiar with their characters and the world. If they want to go shopping or argue among themselves (in character, that is) rather than investigate abandoned castles, let them. Time spent developing the characters' personalities is never wasted. Soomer or letar they'll probably grow bored with what they're doing, and you can nudge them into the plot.
Part 3: Ordinary Play.