One thing that people don't seem to grasp when learning 3d is 'holds'. A hold is just that, you keyframe your character, but you have it stay at that position for a bit before moving on. If you don't use holds, your character will look like he's moving in slow motion, in mollasses, on valium. And no-one wants that.

Say for example, you were syncing to, oh I don't know, say Korns freak on a leash.
The first line I was working with goes "bOOomm Ak dA mMMmmm" (that's directly from the lyric sheet people, I wouldn't make this up). If you only set keys on b, o, a, k, d, a, and m, you'll say 'well that was easy, I've covered all the syllables', but it'll suck. Why?

Because of interpolation between your keys. Take the 'k' sound for example. You drag the 'eee-ooo' slider back so that his lips are pulled back. It looks fine at that keyframe, but if you scrub the frames before it, you'll see that he'll slowly go from slack at frame 0, to 'eee' at frame x. You need to explicity tell AM to not move the slider at all for most of the lead up, and quickly ramp it up just before we hit the 'k' sound. Animating pose sliders (and bones, and everything else for that matter) isn't just about telling the software when a slider should be 'on', you have to tell it when to be 'off' too. Otherwise AM will smoothly move between the extreme keys you set, and everything will fall apart.

Luckily, this is simple to work around, you've just got to be a bit more aware of what's going on.

Thinking about the 'k' sound again, try and visualise when your cheek muscles start pulling, when they hit their extreme, when they finish the hold, and when they are fully relaxed. In an average phoneme action, the muscles would be slack at 0. They would still be slack at 2. Muscles start pulling at 4, hit their extreme at 6, hold until 10, and are relaxed at 14. Fast in, hold, medium fast out.

So to do that:

At frame 0, key your slider at 0%.
frame 2, slider at 0%
frame 4, slider at 90%
frame 10, slider at 90%
frame 14, slider at 0%

This means that when you layer additional slider keys over the top of this, they will interfere as little as possible.

What you'll find however is even if you start and end each phoneme action with 0%, you'll find your sliders will 'drift' slightly between keys. Again, this is interpolation at work. Opening a channel, this is what you'll see:

AM will try and draw a curve that passes through all your points smoothly, which is nice, but in doing so it causes your slider to warble beyond 0, which is bad.

To fix this, go to the first 0 value, and copy the frame. Move one frame into the area that is meant to be flat, and paste the frame. Do the same at the other end. You'll get something like this:

This will 'peg' the spline down, and it won't drift. Armando calls it double keying, and it's a major time saver. Note that if you were to do it the 'Right Way', you'd change the interpolation type of those keys to zero-slope or linear, so that AM will draw a horizontal line through the key. But when you're working, double keying is generally a faster way to work. Set a key, ctrl-c (copy), = (forward one frame), ctrl-v (paste), done.

You may've also noticed that the channel drifts at the top of the curve too. Generally though, that's ok. It adds a bit of a 'moving hold', which means that it's not perfectly still, but extends slightly past the key you set. That tiny bit of movement stops your character from looking lifeless during the hold.

'So why can't a moving hold be used during quiet bits?' you ask. Well it can, but generally you don't want it to. As you animate, you'll see where you need moving holds, and where you need to double key. The only way to learn is to practice.

The other thing that's important is moving the head itself. Compare this avi with no head movement, versus this one. After watching the 2nd one, the first looks almost robotic. A lot of animators will actually concentrate on moving the body first, then the head, and finally do lip sync, to ensure that they get as much character and emotion through body language as possible. You should be able to get a feel for what the character is saying and feeling without lip sync.

I however, like to do the sync first, then do head movements after. But that's just me. =)

Why pose sliders and actions in AM are cool

You've slaved away for hours, scrubbing your sound back and forth, juggling keys like it's going out of style, and finally you've got a perfectly synced up action. But then you decide that instead of a big fat guy, you want a swamp creature to be speaking the line instead. Is all your work wasted?

Hardly! You'll see 're-useability' mentioned all the time in AM documentation, and lip sync is no different. When you defined the action, it keeps track of the pose percentages, and the head bone movement (if you moved the head), but that's it. This means that if you create another head and setup the same poses for it, you can can just assign the action to the new model, and hey presto, there's your new character reading the line!

This means you can now have libraries of speech for your animations, which, as long as you stick to the same poses for your characters, can be used by any character you create.

Other stuff

• You're not limited to setting poses at -100% and +100%. If the movement from one pose to another doesn't look quite right (neutral to ooo for example), you can add another key at 50% so you have more control. You can add extra keys again at 25%, 70%, or wherever you want, so that the transistion from one pose to another runs exactly the way you want it.
• Those poses are a suggestion only. Some people create poses based on major syllables, others create poses based on muscle groups. Try both, see what works for you.
• If your character's face will be seen up close, and really emoting, there's a lot more controls you could put in. Jaw side to side, sneers, tongue lolling, eyebrow middle/left/right, nose wrinkling, cheek puffs, lip biting... Keep in mind though, the more poses you have, the more control you get, but it can get tedious if you have to juggle 50 or 60 sliders just to make your character wink.
• Some controls might be better off on a smartskin. Smartskin is essentially the same as a pose slider, but rather than the pose being linked to a slider, it's linked to a bone rotation. For example, take the jaw open - hhmmf pose. That range of motion could be replicated by putting a bone in where the jawbone would be, and assigning the lower part of the face to it. You'd then create a smartskin so that when the jaw is fully open, you can tweak the points so that it looks right, and when you rotate the bone slightly beyond fully closed, you can tweak points again so that the lips look compressed. Now when you rotate the bone, the jaw will smoothly morph between those poses. You could then create another smartskin for the same bone to handle the side to side motion, that way saving yourself another slider.

I hope this helped!