Passenger shuttle
Posted by: v_mulligan
This image was created with ZBrush 3.5 R3, Blender 2.48, Mosaic 0.4, and Aqsis 1.6. Some textures were generated using Darktree Textures 2.51, and some UV layout was done with Maya 7.0.1. Retouching was done in Photoshop CS3.
This scene contained six light sources with 256x256 shadow maps, and two reflection-mapped objects with 512x512 cubic maps, requiring a total of nineteen render passes. The final image was rendered at 1600x1200. Total render time was 4.8 hours on a dual-core 3 GHz Pentium IV with 1 GB RAM running Windows XP SP3.
The full-resolution version may be downloaded from my deviantART page.
Posted by: mohanohi
Great job
Its really a nice render. But 4.8 hours for single frame? OMG!! I think i still need to learn..
Posted by: v_mulligan
Yeah, it'd be faster if Aqsis
Yeah, it'd be faster if Aqsis supported multi-threading. I get the impression, too, that Aqsis could use some additional speed optimizations. It's powerful, but not fast.
I should try a similar render with ray-tracing in mental ray. Maps should be faster, but the reality might be otherwise, here.
Posted by: v_mulligan
Incidentally, a good chunk of
Incidentally, a good chunk of that time was for generating the maps, I think. The time for the actual 1600x1200 render was more reasonable (though still not terrific).
Posted by: lancelet
Very nice!
Is that Pandora? (Sorry, couldn't resist...
)
I really like the shuttle bay; especially those lit panels... they're awesome!
I think there are some "problems" (I'm talking purely my own aesthetic opinion here...) with the bumps on the ship though. The small "lit" divots look very much like windows, and I think they work quite well at this resolution. The un-lit divots look wrong though: like enormous dents rather than darkened windows - especially the ones on the right-hand side of the image. I think it would really help to colour all of those divots much darker, and perhaps use higher-res maps to sharpen up the specular hilights at their edges. I get the impression that the bump maps are really just on the edge of being an acceptable resolution for the rendered size of the ship. OTOH, I really like the detail from the reflection maps on the left-hand side of the ship; the bright spots that are the reflection of the planet.
Posted by: v_mulligan
I agree -- I'm not totally
I agree -- I'm not totally happy with the front of the ship. I think there are two problems: first, my displacement map resolution wasn't quite high enough, and second, I think I need a specularity map in addition to the diffuse and glow maps, so that the windows can have sharper highlights instead of the dull shading of the hull. This would help to darken the unlit windows near a specular highlight.
This was mostly an experiment, so I'm probably not going to go back and change it. Next time, though, I'll put a little more work into the windows.
Posted by: v_mulligan
Actually, it might not be a
Actually, it might not be a displacement map resolution problem so much as a UV mapping problem causing local resolution problems. The windows down the side of the ship look okay, but at the front, it looks like the windows are just a few pixels across in displacement map space. I would need to redo the UV mapping so that I have more of an equal-area map in ZBrush and then regenerate the displacement map. I might also want to switch the filtering on the map back from "gaussian" to "box".
Posted by: c42f
Filtering note
A quick note on the filtering: aqsis doesn't support box filtering of textures at the moment, it only supports EWA which goes under the name "gaussian". However, aqsis-1.6 does have some parameters to the texture() call which attempt to control the sharpness. In particular, you can try adjusting the "trunc" parameter which controls how much of the gaussian filter density (which has infinite support) is truncated.
More precisely, "trunc" is the integral of the normalized filter density which is excluded due to being outside the filter support cutoff, so it can have values between 0 and 1 (not including zero). The default value in aqsis-1.6 is 0.05. Smaller values will give you less aliasing and more blurring while larger values will give the opposite. Note that "trunc" may effect which mip level is chosen and this may also effect the results.
Let me know if you find this feature useful, since it's somewhat experimental and I'm not entirely sure whether it's a good parameter to expose!
BTW, nice render, as usual. I like seeing your stuff :-)
~Chris.
Posted by: v_mulligan
Interesting! I did not know
Interesting! I did not know about that. Sounds like large values of trunc() would give you a box-like filter -- the middle part of a bell curve with the wings chopped off. It's probably one of those features that should be left in there for the occasional few who want to twiddle it. I doubt that most would be inclined to change it from its default setting, though.
Incidentally, what does happen right now when you set the filtering to "box"? I had some textures set to "box", and I thought I saw a difference, but it might have just been my imagination.
Posted by: c42f
box filtering
If you specify the "box" filter type, aqsis silently uses "gaussian" instead. That's in line with the RISpec which IIRC states that any options to texture() which aren't supported are silently ignored.
~Chris
Posted by: v_mulligan
Good to know. Thanks for the
Good to know. Thanks for the information!