ZSpline

In the past 5 weeks I was helping out the programmers however I could, that’s why I didn’t update this blog as regularly as before. In spite of spending less time on making pretty things and more looking at screens full of text, I’m quite happy how the past month turned out.

I’ve been programming (or more like scripting, if a real coder reads this) on and off in the last 15 years (with the first few in high-school) but I only worked on smaller, self contained tools, never on complex systems with other programmers.
In the case of Gavit I still didn’t want to touch the code at all. I figured that system design, documentation, modeling/texturing, game/level design and special effect authoring will be enough on my plate. It will be best if professionals take care of the code side, ones who know the engine inside out so stuff will be done the most efficient way possible.

What I only realized recently is that there is room for me to contribute to the programming tasks, especially when it comes to experimental and engine agnostic solutions.

The programmers I hired live on the other side of the globe and work on many projects at once. When I sleep they work, when I wake up they’re about to leave the office. I have a very limited window to chat with them and they have limited resources to fulfill my wishes. Turnaround time is long and communication is relatively difficult, in spite of the absolute professionalism and total good will of the participants. It’s just an inherent problem with remote work situations.

As the holiday season came up I felt that I can’t avoid chipping-in the programming tasks. I did smaller stuff for the project before but that was just messing around in Notepad++ which was fine but I figured this time around I’ll need proper tools for the job. I installed VisualStudio with nFringe and jumped in.

My productivity instantly increased ten fold.

There are so many things which are basically the waste of the time of a veteran programmer. You don’t need 10 years of experience with Unreal to be able to derive velocity from a series of positions, to smooth mouse input or convert between two coordinate systems. With some common sense and google skills a surprising amount of practical problems can be solved, as they arise.

With a proper IDE having my back the tasks became puzzles to solve. The proper architecture (made by real coders) often guided my hands when I wanted to extend functionality. I knew what I wanted to do, the logical structure made it clear where to put the stuff and after looking up a few examples I knew how to do it properly.

And I can’t stress enough how much help I got and keep getting from James and Sam. They’re patient enough to explain things to me and don’t hesitate to correct my sometimes sloppy programming practices.

All in all, as a technical artist I really enjoy the tremendous freedom casual programming gives me.

We had to cut Gavit’s character related features (for now) but since characters are somewhat important in games and stories I’m looking into alternate means of showing figures and communicating intentions.

The most obvious solution is posing ragdolls, as seen in the following video:

I used the Razer Hydra in both hands to control pairs of reach targets for ragdoll bones: Head and hips, left and right hands, left and right feet. (Those are the red/green wireframe blobs on the video.) The pairs were selected using buttons while the camera was rotated with the thumbstick.

I learned the following lessons during this mini project:

- Input with 6 degrees of freedom is nice although movement perpendicular to the screen is hard to judge.

- Controlling two things at once does not help posing. I ended up focusing on one hand while tried to keep the other in place.

- 6 control points for a human body is the absolute minimum. Controls for shoulder, chest, elbows and knees would’ve made work faster and the result better. (Why didn’t I just copy Motionbuilder…)

- Clay-like, very tight  joints would’ve made things easier but I couldn’t set up the physics asset that way. (For some reason 0,0,0 velocity target with 9999999 force had no effect at all… o_O )

- Actor relative positioning (as opposed to camera relative one) is confusing.

So it seems that ragdoll posing can be helpful provided that the poser rig does the following:
- Select and work with only one control point.
- One hand manipulates control point, other adjusts view.
- Choice between local and camera relative transformations.
(So it would be a poor man’s Motionbuilder.)

Animating a character real-time will need a different, more marionette like setup, which will be the challenge for the next few weeks, after I finished one more prototype in the pipeline: an Arcade Volleyball homage.

I’ve implemented a counter shader in Unreal, similar to the one I made in ShaderFX:

I managed to correct a huge mistake I made back then so now there is no numeric precision related visual degradation present. Float precision still causes a problem but that affects the actual value shown on the counter: at 7 digits the displayed number is occasionally off by 1.

Toward 10 digits the issue becomes much more serious so I’ll need to come up with some sort of a solution. I hope I’ll be able to simplify the math involved (to more operations we perform the more precision we lose) but if nothing else helps then I will just duplicate the tree and process the incoming value in two parts.

The Razer Hydra integration progresses somewhat slowly these days. Position and thumbstick data is handled alright but the rest (rotation, acceleration, velocity) is still not implemented. I’ll need at least rotation to be able to make a minigame where one must bounce a ping-pong ball from one paddle to the other to get points (displayed on a fancy counter).