Topic: 3D Modeling WIP Topic

[ratty redemption]

I tried a slightly different method of shrink wrapping but it won't then decimate the mesh, due to loads of non manifold errors. so tomorrow I'm going to try again. anyway here's an image of today's attempt.



edit 1: I managed to fix the shrink wrapping this images shows the mesh still with the off cut verts which I'll remove when I decimate and clean up the mesh tomorrow.



admin: Ratty where-ever possible try and consolidate posts rather than doing consecutive ones, it just keeps things a bit more tidy and logical. kat

edit 2: here is the cleaned up mesh after skrink wrapping.



I think this is detailed enough and is about 1000 tris for the 1st out of the three cliff sections I'll be working on.

edit 3: after some tweaking of the new mesh to help baking, I've now got a working normal map. I also tested my glsl material from my rocks on this cliff face with a few scale values increased. I'm intending the crack layer to have more cracks, even if it's not going to be used on this model.

[ratty redemption]

sorry kat, and I will try to remember to do so. if I ever forget feel free to merge my posts

[ratty redemption]

starting a new post for this image as the previous post was getting too long I think.

I've not done any more work on the crack normal layer today, but I think I've improved the diffuse and basic/detail normal layers.

in this perspective image, the camera is quite close to the cliff face and there is still a lot of high frequency detail. next on my to do list is try to mix in some more variation to the diffuse and then add to some cracks, over the next few days I'll then work on the other 2 sections of this cliff.



edit 1: here's two images of the latest diffuse tests. I used a photo source of real rocks for the basic diffuse layer and a smaller tiled dirt texture for the detail layer. the normal maps are the same as in the last images. also none of these recent images have ao or spec maps, just the uniform spec lighting that comes with blender.



edit 2: in between mainly doing real life stuff over the last few days, I've managed to do enough of the new rock material to be satisfied for the time being. so for the next few days I can focus on the meshes again. after that I'll probably do some more work on the crack normals.

[kat]

What did you get the final poly count down to? I'm not sure you really need to do too much else to this, based on the shots above it looks pretty good so I wouldn't go OTT with the 'crack' layering else you may end up over-egging the pudding, especially if the cracks don't follow the general contours present in the mesh. Did you use a tiled texture on this as well? That might allow you to do some nice blending top and bottom with other textures to break up the 'uniform' appearance you've got atm. Nice job so far though *thumbsup*

[ratty redemption]

I've not yet tweaked the mesh so thats still about 1000 tris for this section.

the material consists of a 1024 size dark grey photo sourced rock texture, which is tiled 2x vertically and horizontally, then a 256 size light brown dirt texture tiled 24x, a 1024 size baked normalmap from the model mixed by nodes with a baked normalmap of the diffuse, then a 1024 size detail normal map with cloud type noise, and finally a specmap mixed in the nodes from the diffuse.

so no specific crack detail layers, do you really think I could leave them out?

and can you elaborate on the "blending top and bottom with other textures" do you mean rock to grass type alpha blending?

I'm open to any suggestions for improving this but agree there probably isn't much more needed.

[kat]

The problem with cracks is the same as we've discussed before in that they have to follow a basic principle regards to *where* they appear. The problem most people have with rocks in general is that they don't think about what it is they're doing, by that I mean they don't think about the *type* of material the object they're working on is made of - rock breaks in a particular way, and different rock breaks in different ways. So cracks and stress fractures appear in specific places for specific reason, so whilst you could do something very generalised there's also a danger of over doing it and removing the 'reality' all objects need to have for us to 'believe' what we're observing because the tiling places the cracks in odd places that don't make sense relative to how we understand rock to behave. So yes, based on the images you've posted, adding cracks may not add anything in terms of believability.

Instead, what I'd look at now is breaking the uniform appearance by now using some vertex blending, run some grass along the bottom or top or the rock face so the object looks like it 'lives' somewhere. You might be a bit restricted here though depending on where the models are going (engine tech) as you'll be limited to what sort of blending you can do (the same problem I've got with those terrain sections I was working on for Unearthly 09).

[ratty redemption]

understood kat and thanks a lot

[silicone_milk]

How does the shrink wrap process work the way you're using it? Are you forming the shape with the rock particles and then taking a super sub-divided mesh plane and shrinking it around the contours of the rock shapes to get a bumpy terrain mesh as a result?

[ratty redemption]

milk, yes that's basically correct

imo the advantage of this workflow as opposed to modeling the low poly mesh by hand from the beginning is that once the particle library has been built, which is somewhat time consuming, the library objects can be reused countless times and the particle painting is very fast. also we don't have to worry about the poly count of that stage, or the edgeflow, or the particle objects overlapping.

after a high density mesh has then been shrink wrapped over the particles, they can be discarded, although the original library objects are likely to be kept to be used on another model. and the new mesh will need some cleaning up by hand after its been decimated down to the desired poly count.

I guess it's an alternative to high poly sculpting, but for me being slightly more technical then artistic its a lot easier to produce desired results. and a less powerful computer is needed as the most processor intensive stage is the decimating, which can take several minutes, but surprisingly the shrink wrapping is almost instant and because the high poly mesh doesn't need editing by hand, blender seems to be able to cope with it quite well.

[silicone_milk]

@ Ratty - Cool beans. I imagine you could apply your shrink wrap technique to some spherical shapes and get some wicked organic stuff going on (thin film with a fresnel shader over something like a bunch of eggs or something)

[ratty redemption]

milk, good idea and yes, plus with recent work for dersaidin, I've found we use non planar meshes for the shrink wrapping, in my previous tests I could never get those to work, so in theory any shape could be achieved.

[dersaidin]

Hey guys,

I'm looking forward to see Ratty's cliffs ingame, they're looking great so far

The limit of 5000 was for the entire cliff (2 other sections), and in retrospect it is rather large.
Under a thousand for each section will be more appropriate, still quite generous. However, these particular cliffs are in a position where vis portals will be able to exclude them quite effectively when they're out of sight, so we can afford this tri count.

Heres a short video (30MB) showing some spell casting and some of the building models Trak has done. This is the same map that Ratty's cliffs is for. As you can see the HUD, spell effects, and most of the map is still certainly WIP, but its starting to come together. Theres a few graphics artifacts there (glsl water shader = noisy pixels) due to recording to video.

[ratty redemption]

I've not done any blender work for a couple of days but as I told dersaidin in an email, I managed to get the tris count of the first cliff section down to 700, any more then that and I think it would loose too much of its boulder type shapes. for the next 2 sections I plan to have some smoother parts so they can be optimized more.

heres a series of images with various materials layers turned on.

Heres a short video (30MB) showing some spell casting and some of the building models Trak has done.

the buildings look good so far, but I was getting some motion sickness watching you wander around and especially side stepping while casting the spells, due to the view bobbing being on. can it be turned off like in d3?

also sorry to say but that water looked awful and was also giving me a headache, surely that can't be the result of the capture app or codec? I seriously would leave that out of any videos until you guys have got it smoother looking.

[silicone_milk]

I managed to get the tris count of the first cliff section down to 700, any more then that and I think it would loose too much of its boulder type shapes. for the next 2 sections I plan to have some smoother parts so they can be optimized more.

*cough* parallax *cough*

Would it be possible to get some normal/diffuse/specular mapped shots with a point light or two in the render? The images are a bit dark on my monitor so I can't really make out any features on the textured shots.

[ratty redemption]

@ milk, I'll take some better lit shots later

and elaborate please on what you were thinking with parallax? since d3 doesn't support it natively my only experience with that effect is in crazybump.