need help with mobile asset pipeline

BelgianBoolean

Hey all,

I've been working on a model using mobile (i.e. iphone) specs to try and learn some of the techniques. Truth be told, I've found myself a wee bit lost. My idea was to make a high poly and low poly just to get an occlusion and cavity pass in xnormal as base textures. Is this a proper workflow? Are their other options? The result of my baking has been a mess. All the edges are popping up even though I smoothed normals on it.

[/URL]

Ace-Angel

Without seeing the model, it's hard to say, however, it looks like a triangulation issue. Did you try exporting the models out as quads in OBJ format and baking those? Same result? Are both the cavity AND AO like that?

Also, how are your smoothing groups set?

Justin Meisse

I'd skip the whole high poly part, that's total overkill for what will end up as a tiny model on a tiny screen with a texture around 256 x 256 or less. Paint it by hand.

butt_sahib

yes, handpainting it is definitely the better option.
However, it seems some people find making a high>low occ method useful, so it just depends on how yiou go about your pipeline.

CheeseOnToast

Try setting the baking method to "cosine" instead of "uniform" on a low poly asset like this.

BelgianBoolean

Hey everyone! Thanks so much for your responses. Justin was right when he said it was overkill. My modeling skills are stronger than my texturing, so I was hoping to cut corners. I never was able to solve that odd triangulation effect, so I hand painted everything in Max's viewport canvas. And it worked perfectly! Client was super happy.

@Cheese May I ask what the difference is between those baking settings?

@Ace The model was all quads to begin with, and I double checked my smoothing groups. It's bizarre. I've baked stuff more times than I care to count, but never ran into anything like this. All my maps had the same distortion.

Ace-Angel

The Xnormal help document will tell you everything you need to know (although it's slightly outdated.)

● Uniform. Points are sampled evenly. Usually you will get very pronounced results with this option.
● Cosine. Points are distributed using a cosine probability function ( so we get more samples near the
normal ). You will get smoother results with this option.

The reason I asked for the groups is because the ONLY map which gave me such issues was the Height Map, and it already says in the Xnormal Doc's that you need to divide your model at least 2 times before baking the Height Map.

However, in the case of AO, this shouldn't be happening. Since you said all your maps have distortion, I'm think there is a problem in the exportation process.

Care to post some Screenshot and tell us what format you exporting in?

BelgianBoolean

Hey Ace, thanks a ton for the info! It's pretty enlightening. Unfortunately, I only had permission to post the map from my work (I'm under NDA); and not the model. I did however run some tests just for curiosity. It seems that the issue was that I didn't uncheck the optimize boxes in Max's export. Those seemed to be overriding the normals I had set.