Now that I have Poser, what I'd really like to do is figure out how to use it along with Flash to make interactive animations.

Now there's a stupid way to do this, which is easy. You figure out ahead of time all the things you're going to allow, a very short list, and then you animate them using Poser, render the lot into individual frames, read all those frames into Flash, and use it as a frame-switcher with a tiny amount of logic on top to control what plays, when. The result can look damned good, as long as you don't try to resize it. But the resulting SWF file will be mammoth.

The second way is to do that same work, but to look at the finished images and to cut out the parts that move. Then you use a base background, and make smaller sprites which you place on top of that background. So for instance, if you want a figure which is standing perfectly still but whose eyes are blinking, you can render with the eyes open, again with the eyes shut, and use a graphics editor (or other tool) to cut out just the little bitty bit on the face which contains the closed eyes. Then place that on top of the figure periodically.

That can be quite elaborate, and it's a lot of work. But it has virtues, if you're careful. For one thing, it means that multiple animated parts can be doing what they want to do asynchronously, which is impossible for the first approach.

If you want to get really fancy, you actually use Poser itself to create those parts. It turns out that if you render against the "background color" and store the file out as a PNG, it creates an alpha channel. It looks amazingly good, too, when composited on top of something else:

I composited that with Paint Shop Pro (on top of a picture I stole from J), but Flash can do that at runtime.

I can animate a scene, and then render it multiple times. For instance, if I want the figure's arm to move, I can turn off everything else. What will show up in the render is a disembodied arm and hand, which will do all the things I told it to do, and the render against the neutral background will include an alpha channel for the lot. Trim that, read it into Flash, and now you're really cooking.

But the resulting SWF will still be immense. If you're doing anything at all complicated, you're talking multiple megabytes. And in any case you're using pixels, so you can't scale anything. And if you want to rotate anything, you have to do it in the render. Rotated bitmaps look terrible.

I've become more and more impressed with using stroke graphics for things. It renders quickly if you don't do anything really weird, and it's infinitely scalable. The Guchuko flash I did one time came in at only 6.5K for the SWF file. That same SWF done as a bitmap would have come in at upwards of 100K and it wouldn't have looked as good.

So here's what I really want to do: use Poser to create images, and then convert them to vector images. And that's a real bitch, especially with the tools I have.

Flash has a tool which can convert an image into a vector representation. But if you don't process your image heavily, it doesn't do a very good job. In practice what it does is to turn each pixel into a vector square. So it "scales" but it looks just like the bitmap does scaled, and it's even larger, and renders gawdawful slowly at run time.

So... what you do is to threshold the image. Like this:

Then you can read that into Flash and tell it to vectorize it, and this time the result is quite clean and small. All those big areas with sharp edges? It eats those right up, if you use the right parameters on the conversion.

I did it, too.

(The flickering around the edges of certain areas is a Poser rendering issue having to do with lighting. I know how to prevent that, but this was a test run and it wasn't worth going back to do it again.)

That resizes cleanly, and it's only 63K. But it was really painful to produce that. For one thing, the vectorize option in Flash can't be batched, so I had to manually convert every frame. For the 19 discrete frames you see here, that was annoying. If I had a couple of hundred, it would be out of the question.

What I need is something that will threshold and vectorize a bitmap image in a single pass, which I can somehow use to batch process a swarm of images all at once (via a shellscript if no other way). In other words, I need Potrace.

Or rather, I need Potrace which is smart enough to handle color, multiple colors at once. Probably Photoshop can do this (Photoshop can do anything) but I can't afford to buy it, and I won't steal it.

Bummer.

Posted by: Steven Den Beste in at 06:23 PM | Comments (11) | Add Comment
Post contains 868 words, total size 5 kb.