What began as a frenzied effort to make a bunch of characters that I can use in crowd scenes for both The Lair of the Clockwork Book and The Toaster With TWO BRAINS has unexpectedly turned into a whole new character creation process and, along the way, a whole new generation of character models. I’m excited about this even though I continue to worry about my Thrilling Tales schedule.

Because the object structure of my male heads is different this has meant repeating a lot of the setup work I’ve done for my women. But I’ve also done something new and neat that nearly works as planned – more on that below.

It started during the summer with tasks that I sprinkled in between my Clockwork Book illustrations. I created a series of very low resolution clothing objects and used Mudbox to sculpt them into much higher resolution models. Then I exported them in high and, um, less high versions. I completed the materials and textures on the high resolution models, then baked that texturing and the higher res normals into my medium res objects. (I did this in two different image resolutions, so I could switch between textures depending on how large a character appears in a scene).

Since then I’ve been working on new character heads that use morph targets for their expressions. More importantly, though, I’ve been working on the process for creating those heads so that it’ll be much easier to make new ones now, and the results should be much better.

A morphing character head consists of a base object – the head in a neutral state – and a bunch of other variations on it with expressions or other changes. So long as the structure of all these heads is exactly the same you can dial up one expression or another on the neutral head. In effect you have a single head that you can vary as needed.

Some of the things I’m doing involve changes to the objects. The simplest of those is lowering their resolution. Now, making changes to morphing objects is extremely tricky. You have two choices: either apply those changes to the neutral head after the morphing modifier – which is perfectly safe – or apply the exact same changes to every version of the head before the morphing modifier. Some modeling changes just can’t ever work on the raw targets. Some work most of the time. Some seem to work all of the time. As I work on my modifiers, I have to test them repeatedly. And it’s also important that these changes will work in the same way on every new head I make.

But the results are worth it: ideally, I can create any male character head and apply the exact same changes to it. After all that initial work what I have is a much more streamlined system to use over and over again.

Like I said, polygon reduction was the main thing I was interested in – but as of this morning I have a usable system for doing something completely different.

Switching to morphing heads makes some things more difficult. In the past I designed a style of headgear that’s very common in Retropolis: a kind of form-fitting cowl, like an aviator’s helmet, that hugs the head and wraps around the chin.

When my characters were moving their heads and faces through bones, this was no problem. The cowl would be skinned to the same bones as the face and so it would move as the face moved. But the switch to morph targets meant that couldn’t work any more. The entire face would be moving in a way that the cowl couldn’t follow automatically.

So what I set out to do was to create an object modifier that would cause the cowl to grow out of the head. However the head moved, the cowl would grow out correctly because it would always be based on the current state of the head – and remember, that ought to be any head. By doing it this way I could apply the cowl to all the morph targets, if I liked – they’d still have an identical structure once this same change was applied to all of them. But along the way I thought of something much smarter.

By adding the cowl modifier after the morph takes place I could have my morphing character with headgear; if I turn off that modifier, though, I have the exact same character without the headgear. Slap some hair on the head and I’ve got my guy both ways.

And it almost works perfectly. I’ve been referring to my cowl-growing process as "a modifier", but in fact it’s two modifiers that get applied in sequence. There’s still something wrong when I apply the second one – I’m finding that I have to finish the process by hand, once, for each new character. But even if I can’t sort out that problem I have a nearly automatic process for growing a Retropolitan aviator’s cowl right out of any new male head I make.

Sometimes, I feel really smart. This is almost one of those times.

Now I just need to crank out a few more of these guys before I have to go back to work on the last of the illustrations for The Lair of the Clockwork Book. There aren’t as many to do in this batch, but I know they’re going to be difficult ones. So, you know, I worry.

Well, it’s not really all about the ladies; but they’re what I had to drop into this picture.

During this round in which (theoretically) I’m working on Part Two of The Toaster With TWO BRAINS (while also catching up on a web project) and in which (actually) I’m building character assets like mad because I need so many folks in the background for upcoming scenes not only in TWO BRAINS but also in The Lair of the Clockwork Book, while still, truly, catching up on that web project, and – as I see with some dismay – constructing overlong sentences that just won’t die, I have been practically incommunicado because my nose is pressed so firmly into my monitor.

But I’ve done great, though undetectable, things: I’ve completely revamped my work flow for creating female character heads, with much greater controls over their expressions, and every one of these new heads uses the same object topology so that if I really wanted to, I could probably morph them from one character to another. This involved some difficult work over five days – not counting making the actual heads you see here – but it’s the kind of thing that will continue to help me out in the future. The new system worked so well that I even rebuilt my Gwen Hopkins character’s head completely from scratch, since she has a lot of upcoming scenes in TWO BRAINS.

Male heads, on the other hand, are still ahead of me; I have to redo that preliminary work for them because their object topology will not be the same. But happily I still have the original (though incompatible) data for Nat Gonella’s head, and I’ll eventually be rebuilding that one with the new system, too.

But it’s not all about the heads. I’ve also finally done some 3D sculpting work that I actually like for clothing. I’ve turned out a bunch of different base garments and a number of texture variations on them, with more to come, so that you won’t think everyone in Retropolis shares the same clothes. This new approach involves high and low resolution versions of the garments, where I’ve built them completely in the higher resolution and then baked their textures and normals into low resolution versions of the same clothing. I’ll even have the higher resolution clothes there for use in some of my really large renderings.

So this period of a few weeks is just about all concerned with what, in a perfect world, would have been pre-production. I just don’t have the luxury of doing all that work at a more sensible time, when I’m not still under the evil eye of my schedule. Because for as long as I’m doing a serial I am always under the schedule’s evil eye.

The bothersome thing is that this work – though it will benefit TWO BRAINS – is coming out of the TWO BRAINS part of my schedule. The publication schedule for the Clockwork Book makes it a cruel master.

I just need two of me, I guess.

Anyway, it’s back to more of the same and some other stuff. I ought to get back to The Clockwork Book in a month – it would be better to do that even sooner. In the meantime, I’m an asset building machine.

Made as the filmmakers’ final project at the Utrecht School of the Arts, Blik is a story told entirely without dialogue and – because the characters’ faces lack any features – without the use of any expressions at all. The whole thing is done with an exacting use of body language and as challenging a prospect as that is, these folks (now at Polder Animation) pull it off brilliantly.

There’s something very pleasant about their combination of watercolor-like rendering and their use of light, shadow, and – at one point – rain. Very nice work!

Next week, Autodesk will be releasing its free photo to 3D application Photofly. This video shows how it works. You’ll see they’re getting pretty respectable results without a sophisticated camera setup, so that it’s possible to take all the photographs for a building model by simply walking around the building, or of a head by moving around the person and taking pictures from a variety of angles without any kind of carousel for the subject or a fixed array of cameras.

The system uses a local Windows application called the Photo Scene Editor to assemble all the reference pictures and submit them to Autodesk’s servers for processing. Within the editor you can make some changes to your geometry based on entered measurements. Some amount of cleanup is also possible in the Editor. You can see more about the Editor’s features in the video below.

Any system for 3D digitizing gives you some crazy topology that has to be cleaned up. In earlier studio systems this was always combined with way too much geometry; here you can see that the triangular polygons, whose numbers are comparatively restrained, are still constructed with a bit of eccentricity.

The environmental models could probably be used as-is: the same goes for static props. But for character work you’d really need to retopologize the model so it’d animate well.

As I’ve watched these (and some other demos), I’ve been thinking about how well this might work for digitizing costume bits, clothing, and even plants – naturally draped and shaped objects that are difficult to create with polygon modeling. I think the possibilities for clothing and costume parts are especially interesting, given how many steampunk fans go completely mad with their costume work. Neat!

There’s recently been a lot of interest in adapting Microsoft’s Kinect into homebrewed 3D digitizing systems: but this, which is first of all free, and secondly requires only a digital camera, looks awfully promising. I guess we’ll find out in a week or so at the Photofly site.

Technorati Tags: 3d scanner, 3d digitizer, 3d modeling, autodesk, photofly, computer graphics