Rather than trying to process Poser renders, I thought it would be interesting to try to vectorize some frame grabs.

So, I grabbed my copy of Grenadier and went through it, and finally found seven appropriate frames in episode 5. Rushuna bows and says, yoroshiku. Here's an animation based on the direct frame grabs:

I loaded the first of those into Inkscape and started messing around with the conversion parameters. The first thing I found out is that you have to use "stack layers". If you don't, then the resulting file looks like crap when it's loaded into Flash. The bad news is that "stack layers" about triples the number of nodes.

Anyway, after long experimentation, here's what I ended up with for conversion parameters:

Multiple Scan mode: colors
Number of scans: 16
Smooth = no
Stack scans = yes
Remove background = no
Suppress speckles = yes, size = 2
Smooth corners = yes, size = 1
Optimize paths = yes, tolorance = 0.2

And after manually converting all 7 frames, a process taking about 15 manual operations per frame, and then making a flash animation out of them, this is the result:

It looks pretty good. But it's nearly three times as large!

The one based on the bitmaps is 201K. The vectorized one is 537K. I'd say this experiment is a failure.

To make the result smaller, I'd have to drastically limit the number of planes, and it would look dreadful. Even with 16 planes (which means 16 colors) it doesn't look as good as the original.

I think the problem is that the background is too complex. The cels were created digitally and use a small number of colors, but the background is painted. That's quite common these days. I think they like it because it makes the end result look richer without making the animation job prohibitive.

I think I might look to see if I have something with uncomplex backgrounds which might ameliorate this a bit. But I doubt it will help. I suspect the only way this would ever really look good is if the images were hand traced -- and that takes hours per frame.

UPDATE: Or I could go get some of those frames of Munemune where she's shown against a white background near the end of Magical Shopping Arcade Abenobashi. Of course, those are dreadfully NSFW. (This is bad?)

Posted by: Steven Den Beste in Flash at 10:46 PM | Comments (2) | Add Comment
Post contains 402 words, total size 3 kb.