What's the state of Iray development?
Sevrin
Posts: 6,307
in The Commons
I visit the Iray Dev Blog now and then, and there hasn't been an update in over three months. You don't see Nvidia mention Iray ever in their newsletters. And now, with the last and current iterations of Iray, we have this opacity problem. Is this the end of the road for Iray or are they working on something so mindblowing that they can't talk about it? I guess we can hope it's the latter, but I'm concerned it's the former.
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
I'm pretty certain Iray development will continue — especially since Iray is integrated into NVIDIA Omniverse (which the entire company has been heads down on for quite a while).
You should definitely check out GTC 2021 this month for more details, it's free, and the keynote is April 12th at 8:30am PDT...
https://www.nvidia.com/en-us/gtc/
Yeah, I get their newsletters, but I'm not going to listen to Jensen wade through hours of AI/ML on the off-chance to see an Iray logo buried among a dozen others on a Powerpoint slide for 5 seconds. There are a couple of sessions that mention Iray, so perhaps it's not completely dead. It's just that Iray got talked about a lot more by Nvidia 4-5 years ago than it is today.
I have no idea what I would *want* from iRay development, other than optimizations.
What I was really hoping for was their "2020 hair stuff" making it to DS in some meaningful implementation (hForce, my hopes are fading...)
The iRay hair nativly implemented in DAZ would be nice. I think they were working on toon shaders for a future version of iRay which would add another useful utility to iRay.
I have no idea if Nvidia is working on adding heterogeneous volumes to iRay, but that would be a great addition to a future version since the homogeneous only volumes really limit smoke/water/fog/cloud/fire from realistically being rendered.
The top blog post from December 15, 2020 seems to be pinned — but if you scroll down, the next one under it is from March 23, 2021... for GTC 2021.
There are Iray-specific sessions that will start streaming on April 12th at 10:00 am PDT, UTC-7...
Iray in Omniverse: New Features for Light Transport Simulation and Rendering [S31739]
Learn How NVIDIA Iray Physically-Based Rendering and RTX is Driving Enhancements to Consumer Products and Retail [S31751]
SOLIDWORKS Visualize: Recent Developments [S31988]
Rendering Realistic Figures: A Detailed Look at Skin Shaders Using Iray and MDL [S31893]
Related sessions:
Sharing Physically Based Materials Between Renderers with MDL [S31207]
Integrating the NVIDIA Material Definition Language in Your Application [S31241]
OptiX Advanced Topics [S31752]
What's New in OptiX [S31736]
What is the opacity problem? I've been distracted by IRL last few months and am behind on things.
It's slow, it's clunky, the materials are weird, it's CUDA limited, and pretty much nobody actually uses it outside of Daz.
In otherwords, it's basically same as its always been...
The iray dev blog now has links to the "what's new" GTC presentation (to watch it a GTC registration is required).
https://blog.irayrender.com/post/646465054674403328/gtc-2021-sessions-containing-iray-tech-you-dont
I just had a look. I found interesting some information regarding volumes. Seems so support for inhomogenous volumes is coming - hopefully also to DS in the end.
Heterogenous volumes would be a great addition! I wonder how they would be represented. VDBs would be a logical choice. Since DAZ Studio doesn't have any volume primitives I would think that they'd have to add support for them first before they can add support for the Iray heterogenous volume feature.
they should skip that loosy Iray and put Blender Cycles into Daz 3D instead
much more realistic lighting and shaders.
...didn't poser sort of do that about five years ago with their Superfly render engine that is based on Cycles on response to Daz adopting Iray?
They posted some screenshots of the presentations including one on the new volumes:
https://blog.irayrender.com/post/648366479710076928/highlights-from-the-gtc-iray-in-omniverse
I think it might be faster too, at least there seem to be more options to optimize the scene without making everything look terrible.
Plus, then they could replace Filament with Eevee. I was not impressed with Filament.
...as long as I don't need to deal with the "overturned bowl of spaghetti" that is the Shader Mixer to convert Iray materials to Cycles, I'd be fine with that. I am pleased by the fact that it has out of core rendering like Octane, but without the subscription hassle.
Blender's Cycles renderer is great, and has a bunch of CUDA and OptiX optimizations for NVIDIA GPUs including optix hybrid rendering (GPU and CPU) in 2.92.
There's a bunch of good sessions on Cycles rendering in the NVIDIA STUDIO youtube channel... I like this one from Ben Mauro:
Real-Time RTX GPU Rendering in Blender 2.8 Cycles/EEVEE with Ben Mauro | NVIDIA Studio Sessions
I'd love to be able to use the Cycles Renderer directly in DAZ Studio; but for now, the Daz to Blender Bridge is a nice alternative to get DAZ 3D content rendered in Cycles.
How bad is the workflow to convert Iray materials to Cycles?
I wonder if there is a better way...
Not legally. Cycles specifically has a different liscence than the rest of blender that allows it to be used in closed source programs - Eevee however has the same liscence as the rest of blender as it is to integrated to seperate in the way Cycles was.
Also to the original point, as someone who uses both the idea that Cycles has more intrinsically realistic shaders or lighting is... misguided (lights in particular. In blender its hard to even attempt using any real world units at all, which personally I would consider an element of "realistic") Cycles does have a dedicated SSS shader rather than relying on volumes, but even this isn't more "realistic" just way less of a pita to set up. Most of Cycles real advantages are being in blender - hey intancing and particles work better in cycles, (because blender handles those better,) using nodes for you shaders is so much easier for cycles (because blender will update them in real time) etc. If you take cycles out of that environment a lot of those benefits go away.
In my experience Cycles in more memory friendly but again its hard to say how much of that is Cycles and how much of that is blender. Blender has things like adaptive sampling that can make renders faster - although the user has to know what they're doing - theres no one size fits all settings to speed up renders. Iray might actually be a bit faster at interiors lit from the outside, though both are not great and luxrender knocks both out of the water on this one.
Cycles is great but its not something that Daz could just plop in and woo everyones renders get better because its better.
That must be how Poser was able to add the Cycles rendering engine, but not Eevee... Superfly is just an implementation of Blender's Cycles.
Filament seems like it has a long way to go — I too don't like the look of its renders. Its documentation on physically based rendering is an academic deep dive on the subject...
https://google.github.io/filament/Filament.md.html
Eeevee is cool, but it also has a lot of limitations...
https://docs.blender.org/manual/en/latest/render/eevee/limitations.html
Here are some really good resources on real-time rendering...
https://www.realtimerendering.com/index.html
Iray has so many flaws that it would be nice to finally see some day Big names (VRay, RedShift) develop a paid plugin for DAZ. Amazing that except Corona nobody really want to invest in an engine for DAZ. So many customers would be willing to invest just to get better renders.
We can't add emissive objects that will lighten the scene while being invisible at render. We can't ask a light to ignore some objects. We can't ask a light to calculate shadows for some objects but not for others. We can't calculate a motion bluR ! There's no way to generate fur, grass at render time without using heavy subdivision and a displacement map (weak solution), etc. Many options simply don't exist in Iray. Options that other rendering engines have for many years. Even now dead ones (Final Render, Brazil) had more options than Iray :/
Now that there is Filament, a real-time engine, I'm scared that even less work will be put into developping Iray. Let's hope some huge update will come up some day soon.
None of these are physically possible, so it's not really surprising it's not included.
Iray still have no motion blur. We can't add emissive objects that lighten the scene but are invisible at render. We can't ask a light to ignore some objects. We can't ask a light to calculate shadows for some objects but not for others. There's still no way to generate fur, grass at render time without using heavy subdivision and a displacement map (weak solution), etc. Many options simply don't exist when other rendering engines have them for many years. Even dead ones (Final Render, Brazil) offered more customization of the render than Iray does.
NVidia is investing way more in the future of AI than in GPU based rendering.
But Iray 2021.xx versions are coming. So let's hope there will be some major improvements.
But when you see those kind of Genesis characters rendered in Blender... you realize how far behind is Iray ;)
Artist : Blitter.
Set the emissive surface's Cutout Opacity to a very small, but non-zero value.
Light Path Expressions through canvasses, in the Advanced tab of Render Settings.
There's another store with plenty of blue motions. If you mean motion blur, that is a legitimate complaint.
Strand based hair should be able to do some of this, or dForce hair in store products, or Look at My Hair. Iray hs gained its own fur system, though dForce/strand based hair doesn't yet use it - I'm not sure if that can substitute other geometry for hair strands though.
Neither Filament nor Iraya re developed by Daz, so how coudl one take resources from the other? In any event, Filament is presented as a drawstyle for previewing rather than as a final render option (though soem people uise it that way, and some PAs have released products to help them)
Iray has motion blur, but DAZ has not implemented it in DAZ Studio https://www.migenius.com/doc/realityserver/latest/resources/general/iray/manual/index.html#/concept/motion_blur.html
Iray also has a native fur/hair that could be used for grass but again, it has not been implemented by DAZ in DAZ Studio.https://blog.irayrender.com/post/190895183971/feature-time-part-ii-iray-2020-introduces-native
https://www.migenius.com/articles/realityserver6
For what it's worth, Blitter's Iray renders are even more realistic. They are fantastic at texturing, and their particular style hides any issues.
Ha!!!
Theres also instancing which Iray handles just fine (the actual bottleneck is more creating the instances which other programs do tend to handle a bit better)
but how much more grass do you need?
I think thats enough grass personally.
in terms of hair and fur
I've managed to fit quite a bit in one scene. (both the hair and the coat are strand hair - I think the coat qualifies as "fur")
point being there is zero reason to use displacement for fur grass or whatever.
and an example of their Iray work
I think its fair to say the realism of both is much more a testament to their skill than the relative merits of either engine. :)
Is there any news? Both Daz development, and IRay both seem to stalled. Daz support did say that a "Big announcement is due about Daz". I'm really hoping it's not just about some AI stuff.
Some weeks (months now?) ago, they did voting on the forum for the new splashscreen for DS, featuring new volumetrics that can have luminance. The voting ended, no results were announced, and nothing since then... Maybe the big announcement is that? Especially that some Daz Store products already offer those functionalities (with the caveat you have to have beta build to use them)..
The change log shows activity, so I am not sure how you can say it has stalled. Daz has said they will not update iray during a cycle, as the risk of breakage is too great - my undertanding was that an Iray update will have to wait until the next 4.x (or 5.x if that is what we get) development round.