Archive for March, 2011

Gavit Dev Diary #6

Phew, the Necromancer’s staff is finally done:

The far left shows the high definition mesh and textures in 3ds Max 2011′s viewport. It maintained ~30 fps in spite of the object being 770k polygons and using 28 textures with a combined resolution of 45k by 45k.

The second image is the MentalRay rendering. It is still necessary to do from time to time because the viewport can’t show the effect of bump or displacement mapping (or use any other UV set other than the first one due to a bug).
I spent considerable amount of time tweaking displacement values so I didn’t run out of memory during baking… 4Gb is just not enough anymore.

The last two pictures are from Unreal, showing the low poly mesh (7000 polys mapped with four 2k textures). They are pretty close to the original although there is a slight loss of saturation for some reason.

So that’s it. The workflow got tested pretty thoroughly and now I’m sure that the idea is worth pursuing. Adjusting the surface properties with instant feedback is just awesome… it encourages experimentation and following design changes is fast. Of course some pre-processing is required (baking ambient occlusion, convexity and concavity onto the high definition mesh) but that can be automated.

Unfortunately there are a few problems which degrade the user experience:

- After a certain shader complexity, the slate editor slows down Max’s UI quite much. For example in this scene selecting a mesh takes around one minute if the slate editor is open.

- The real time viewport can’t use multiple UV sets, everything is mapped with the first one. (Rendered images are fine.)

- Before starting the final object bake certain nodes must be manually linked for every texture used.

- The final object baking can use huge amounts of memory if displacement is enabled.

I decided to hold off with the tutorial on all this and see what advances/bug fixes Max 2012 will bring in this regard.

This week I’ll focus on the design document and help our recently found animator with the unreal side of things.

Written on March 28th, 2011
Categories: Gavit, My projects
Tags: baking, workflow

Gavit Dev Diary #5

On Monday I started texturing the high definition mesh of the Necromancer staff, using the established workflow: file link the HD models in 3ds Max 2011, create surface node network in the slate editor, render to see what I have. It wasn’t a particularly slow process, Mental Ray is almost as fast as modo’s converging renderer (at similar quality), but every time I used one of my MetaSL utility nodes, I started thinking about how cool it would be to see the high-def surface in real time… But hey, a few weeks ago I spent a considerable amount of time on the MentalMill -> Max workflow and there were just too many usability problems between the two applications.

Yet, I had the feeling that I missed something.

As it turns out I did indeed: a node in the slate editor called “Map to Material”. With this little gem one can make an arbitrarily complex shading network using MetaSL elements and see it in realtime in the viewports. Right in Max, there is no need to use MentalMill. Yay!

But between the first simple test graph and the final, baking compatible workflow lied a 6 days long walk on a technical minefield. The bugs and undocumented “features” provided not a small number of facepalm moments and at the end made some of the procedures a bit rough on the edges. Still the net result is quite satisfactory and who knows, Max 2012 might take things even further.

This week’s efforts will bloom as a tutorial on the topic after I finished the NecroStaff.

Regarding other aspects of the development: the high level storyline of the demo is finalized, so it’s time to rough out the 3d “maps” and brainstorm about the minute by minute events.

I’m still looking for an animator with Motionbuilder experience. My job posting on polycount.com has yielded pings from a web designer company, a flash animation studio and other, less than ideal candidates so I’m still at square one.

The programmer managed to grab his stuff from the shaken up offices in Christchurch, so the code work will hopefully continue next week.

And finally, just so this post too has a picture, here is one of the pre-baked maps used with the high-def object, the convexity map. It encodes three different frequencies in the RGB components and used mainly as a mask for blending stuff like wear and tear.

Written on March 20th, 2011
Categories: Gavit, My projects
Tags: 3ds max, shader, workflow

Gavit Dev Diary #4

Because the programming department will pick up work only next week I decided to spend a bit more time working on the Necromancer’s staff.

The concept was done in a day, mostly using Illustrator.

The staff will be used as a firearm so it needed some kind of a “barrel”, a place to release the projectiles from. The eye sockets of a skull lent themselves for that function. I put the skull on an axle so it can rotate and face forward even when the staff is held horizontally.

At this stage the skull had a long ponytail wrapped around the staff so it felt a bit more stable.

Then I started modeling the high definition model in modo 501. It didn’t go as fast as I expected because at every step I stumbled upon a few bugs. For every bug I double check if it really is one then I try to find a viable workaround and finally submit a bug report with example scenes and reproduction steps. It takes some time but at least I do something about the problems which beats just quietly hoping for fixes.

Anyways, as you can see the design changed somewhat from the original concept: the skull’s ponytail didn’t really work in 3D, I got rid of half the chains so it looks more asymmetric and I exchanged the smaller skulls for a more crooked bottom half.

A rather messy skull object was downloaded from Google’s 3D Warehouse and I used it a reference for a properly made SDS model.

The basic shape of the wooden parts was roughed out using the solid sketch tool while the finer details were sculpted.

Now there is a cloth where the ponytail was wrapped around the wood. I started sculpting the wrinkles on it but soon realized that it’s a slow, tedious and boring way to do this so I looked for a better solution.

I wanted to try out Ryan Clark’s normal map photography method for a while and now it was my chance. I took a desk lamp, a drawing board, my camera and started taking pictures of wrinkled white T-shirt.

I mixed the images in Photoshop to get the normal map then used Ryan’s CrazyBump to convert it to a height map which I could apply to the surface while sculpting.

The results are quite good although next time I’ll try NVIDIA’s Apex cloth plugin for 3ds Max.

As a side quest I finally setup my Amazon S3 based backup system. I tried 8 different clients, both free and commercial ones, but 7 of them didn’t work at all. Not even Amazon’s own upload manager which made me think that my PC is the source of my problems. During the last few weeks I was debugging my system and everything related to my internet access but found no obvious issues.

At the end I settled with Cloudberry S3 Explorer PRO, the only application which was useful most of the time. Manual file management works without a hitch but it does have problems with its folder sync functionality: it fails on certain files for unknown reasons. Fortunately their product support is very forthcoming so we’ve been chasing this bug for a while now, it’s only a matter of time before we nail it.

Next week I finish up the Necromancer’s Staff and since I’m in dire need of a reliable animator I’ll continue searching for one. I hope that the coder guy could make some progress as well and we can finally see how the animations look ingame.

Written on March 13th, 2011
Categories: Gavit, My projects
Tags: amazon s3, concept, illustrator, modo

Bulletstorm multiplayer

Recently I got cable net so for the first time ever I have a good enough internet access for playing shooters online. Not having any other games installed, I gave Bulletstorm’s multiplayer a try and it turned out to be rather fun.

In this mode the game is stripped from all the crap elements (story, characters, conversations, quick-time events) and the game mechanics start to make sense. The Anarchy game mode is strongly cooperative and mildly competitive which ends up being a very tasty mix.

I’m sure that the novelty of online gaming will fade away over time but right now I’m quite entertained by the multiplayer Bulletstorm.
(I also learned that I should read game reviews and watch gameplay videos before (not) buying a game so next time my expectations won’t be off this much regarding the single player mode.)

On a related note, I recommend reading this two part PC Gamer editorial on the topic: Part 1 – Part 2.
Tom Francis’s views are very close to mine, but in his response, PCF’s Creative Director also brings up valid points.

Written on March 8th, 2011
Categories: Games

One sentence reviews #1

Bulletstorm: Interesting environments and great visual design wasted on primitive game mechanics, dumb story and genuinely annoying characters.

Dragon Age II Demo: This demo, featuring butt ugly and totally linear levels, repetitive combat, boring enemies and inconsequential character interactions, will not help selling the full game.

The Walking Dead (TV): A properly written and acted drama which feels fresh in spite of being situated amidst the “done to death” zombie apocalypse scenario.

Written on March 6th, 2011
Categories: Games, TV shows

Gavit Dev Diary #3

This week I fixed the few problems with my MetaSL utility nodes so they are production ready. I also managed to fix a long standing problem regarding baking MetaSL nodes in Max.

This means that one can create very complex realtime shaders (using multiple high definition textures and procedural effects) for the high poly object. Adjusting the visuals and experimenting with stuff is a very fast process as everything is realtime with instant feedback. (Even a slow “realtime” speed like 1 fps is much much faster then re-rendering the object.) Then the final step is baking the final look onto a low poly mesh so the visuals can be reproduced using simpler, more game friendly shaders.

Well that’s the theory at least. Unfortunately there are several usability problems at the moment which make this workflow useless. (For example 3ds Max needs to be restarted every time a shader is modified in MentalMill.) However the goal is not far off and I hope Max will soon gain the ability to produce MetaSL shaders right in its Slate editor.

After seeing a Mudbox video about manually sculpting a rock formation I though I give it a try. However instead of doing manual labor like that poor guy I chose the procedural approach: let’s define the general parameters of the topology and surface texture then let the machine do the grunt work. Here is the result:

This serves as a prototype for making rocks, boulders or cave walls.

Regarding the animated short, its main events are coming together nicely. I have the situations roughed out and this allows me to start planning for the required art assets. As it stands now I’ll need a wide selection of stuff, from medieval ruins to a spaceship cargo-bay. I won’t be able to build everything for myself so I’ll need help: most setpieces will be outsourced while I’ll try to buy the less important background items. I saw a few places where “game ready” assets are sold and some of them looked actually nice. Probably I’ll still need to work with them, tweak the geometry and textures, but I still get a headstart.

I finished the sub-machine gun concept art:

AVC Mp4 9mm

I have three weapons left, a carbine, a rocket launcher and a mage’s staff. I think I’ll continue with the latter and give modern weaponry some rest.

Most of the animations I’ve been putting off are now done and ready for testing. The next thing will be finding a professional animator who I can afford but also does a good job.

Next week I’ll transfer all animations to Unreal and fix any potential problems in the anim workflow.
I’d like to create the mage’s staff from concept to final ingame object in 3 days: 1 for research and concept, 1 for modeling (high and low poly), 1 for texturing and baking. It will help me test the asset authoring workflow and the staff’s generally organic design will allow me to do some sculpting with my underutilized tablet.

Written on March 6th, 2011
Categories: Gavit, My projects
Tags: concept, illustrator