Given that Nvidia is going to offer a Linux version of Iray for rendering server farms, I became interested in it. Sadly, the client-side portion of the rendering engine is not yet available for Linux. A search has not yielded me any answer as to whether Nvidia plans to allow Iray plugins for Blender or not.
Given that, I have been using Daz3d in Windows to do some rendering. It is a pretty nice piece of software given that it abstracts away a lot of implementation details and allows me to focus on the art.
In a previous Windows post, I mention I was having difficulty with HDR maps. I was able to sort that out. I had a couple of things set wrong. First, I had a limited sky-dome setup which was distorting and red-shifting the image. Second, I was rendering at Ultra-HD resolution, which the provided hdr texture would not support. So here's a before and after pic. Note these are full HD walls covered wtfPL, meaning do what you want with them. Post processing was done using the Gimp.
Render times are insane with this new gpu based rendering. These full HD images rendered out to 100% convergence in less than 30 minutes. I did see that Cycles will be coming or is here with gpu acceleration. I will be posting my artwork here:
I wanted to add one last thing: I chose a pretty bad scene to start with, given that the only object in the scene is the character. HDR maps are nice for providing diffuse world-space lighting, however, your point lights won't bounce off of it. This is important in these physics-based rendering engines because they actually calculate the light behavior when drawing. So the best I can come up with is a close approximation of how the light should be falling on the character, given the scene. Forest scenes always have incredibly complex lighting scenarios, and I am thinking the best bet would be to ditch the HDR map, provide a lit skydome for the diffuse lighting, or just use a sky map, then built the forest scene up as actual 3d objects. It would be a lot more work and would drastically increase render times but would be worth the effort in terms of realism.
tknomanzr: Interesting, what is your line of toughts on this renderers?
I mean why give yourself to the mercy of nvidia when there are cycles, luxrender, mitsuba that are free/opensourced (and i guess many more) and also the limit of beauty is very-highly set in this case = user knowhow.
btw: sheep-it render farm does support single images (in 1920x1200px case they are split into 4 tiles which then get propagated into 4 machines, so you can be really fast as well, also this means you can get nice renders from 2009 tech, since you don't really do any final renders on your own hardware < the point being that there is no need to upgrade hardware every year or own the latest nvidia).
p.s. I can tell you that my Maya/vray (or something) friend does hate me silently, thought he never admitted publicly.
example: (this would render about 6-7 hours on my old mac (mostly due to motion blur), but was done in less than 2 on the sheep)
(I didn't count another hour for grade)
This specific scene could be 100% rendered with a gpu (and it might actually be).
pros: commercial people will hate you and a lot more.
cons: cycles is clearly not an unbiased renderer & sheepit doesn't seem to support openexr even for single frames.
Last edited by brontosaurusrex (2015-08-22 14:30:56)
I had to give this one some thought. What I really want is the best of both worlds. I want Blender to be as easy to use as other commercial software. I don't know. I load up Blender and end up totally lost within about 5 minutes. 3D software has a steep enough learning curve as is. I just really don't care much for Blender's interface.
Following up on this line of thought, I was reading on a possible port of Octane to Blender and one of the devs had responded that he wanted too but had to figure out how to create an abstraction layer for the software so that he wasn't linking directly to Linux libraries and thus violating the GPL. I love the concept of free software up until you get the point that people don't get paid to develop it. Sometimes I think the GPL hurts Linux adoption. Good programmers should expect to get paid for their work as should good artists. Not that I am rushing to drop $600 on zbrush atm, however.
Parallel to that line of thought is if you have to charge $600 for software, then at least provide a way for people to use it differently or something. Daz3d abstracted away alot of the low-level 3d stuff, gives their software away, and makes their money off of selling content. It's an interesting business model, at least.
...I want Blender to be as easy to use as other commercial software. I don't know. I load up Blender and end up totally lost within about 5 minutes. 3D software has a steep enough learning curve as is. I just really don't care much for Blender's interface.
I guess it depends on what you start out with. I sort of like the idiosyncracies now!
Regarding development funding, I drop a few $$ to the Blender Foundation quite regularly, because I think it is a fantastic resource.
Interesting render engine comparison on Blenderguru: http://www.blenderguru.com/articles/ren … -vs-giants
I just really don't care much for Blender's interface.
Mastering blender interface complexity must be 100000000x smaller challenge than understanding (any modern) rendering engine, so on the long run not important at all.