Hello everyone!

Finally we are releasing the 0.58 update of CentiLeo for Cinema 4D R16-R21. It was a very complex development with a lot of changes and novel methods. We have got 2x - 10x rendering speedup. The render speed is always at least 2x better than before and depending on the lighting complexity the speedup is even higher. See below!
Some polishing and issue fixing still should be done but results are already very cool!

Download link: http://centileo.com/download/

Change Log: cntlc4d 0.585 alpha (2020 January 27)
[core] 5-10% rendering speedup
[core] Fixed shadow catcher weights in case of many light sources (makes shadows of different intensity per light source on the Matte/Shadow Catcher object).
[core] Fixed shadow artifacts in case of very small area lights.
[core] Bitmap file paths now support Unicode strings (not just English).
[core] Fixed camera Auto Focus in IPR ([AF] button).
[core] CentiLeo updated to CUDA 10.2.
[cntlc4d] Added support for Cinema/OIDN denoiser in R21 for CentiLeo AOVs (see 2 new checkboxes in CentiLeo Render Settings -> AOVs: 1) “Denoise Images”, 2) “Keep raw and denoised AOVs”). If Cinema denoiser is not enabled with Denoiser Effect (which works only for Main Beauty) then CentiLeo settings enable this for it’s AOVs.
Attention! Must Upgrade to:
- Latest NVIDIA Display Drivers: Game Ready Driver 441.87 or Studio Driver 441.66 (slightly faster).
- Minimum Cinema R21 build is 115

Change Log: cntlc4d 0.584 alpha (2019 December 31)
[core] Fixed z-fighting render artifacts for coplanar polygons in the scene, precision is improved.
[core] Fixed few rare crash-bugs.
[cntlc4d] Cinema R21 support.
[cntlc4d] Installation instructions are changed a bit, read them again!

Change Log: cntlc4d 0.581 alpha (2019 November 21)
[core] Fixed bug with multiple lights sampling. Now scenes with many lights have much lower noise than 0.580.
[core] Fixed rendering artifacts in case of rendering out-of-core geometry with near/far clipped camera.
[core] Fixed a couple of other minor issues/artifacts.

Change Log: cntlc4d 0.580 alpha (2019 September 23)
[core] Rendering speedup from 2x to up to 10x:
- Render iterations are around 2x faster almost always (base speedup);
- Render iterations become smarter and reduce the noise better for global illumination, in some cases it is required 2-4x less iterations to get same noise level as before. These improvements are best visible for interiors and night scenes especially for regions with dominated global illumination and lack of direct lighting. This improvement work for all area lights, mesh lights and environment HDRI with portals.
- Mesh lights with complex shapes (more than few polygons) generate 3-4x less noise than before.
- The new algorithm should be always better and more robust than the previous one (less noise per iteration and faster iteration time).
- And there is no cache, no blur, no yet denoiser, no aditional settings, no additional memory consumption

Known issues to be solved (temporarily broken features)
- Dirt shader;
- Absorption property of material;
- Directional/infinite light source;
- Adaptive noise-based image rendering (where pixels are no longer rendered when specified noise level is reached).
- May not initialize if AMD GPU is in your computer together with NVIDIA GPU.
- Very limited material converter tool (diffuse color + texture of C4D material).

Please, test it on you setup and share your renders!
Best regards,
Kirill

Quote
a1cat cato wrote: Hi guys , any updates ..

Hi! Still working , there is very good progress, but still need to increase GPU saturation with tasks, fix few broken features and then ready.

Quote
a1cat cato wrote: Secondly, are those features are only for cinem4D, does the 3dmax version have different features ? What i did understand is that there are 2 version, one for 3dmaxc and the other for cinema4D, every one has different features than the other. am i correct ?

The next release 0.58 will be both for 3ds Max and Cinema, the features will be syncrhonized 1:1. Only node-based GUI for Cinema will be done by 0.59. 3ds Max produce node GUI automatically for all kinds of 3rd party materials. But Cinema developers decided to hide this possibility and make node GUI only for own native materials, so we have to re-implement this basic bicycle.

Thanks guys for good wishes! Will do best for next release.

Thanks I honestly think we also need Random-Walk SSS too for better precision compared to current ray-traced SSS which has few problems on geometry edges.
As for Maya version, I think the initial version with all basic things (not very deep) and IPR may take around 2 months based on our experience with Max and Cinema. And after that will continue never ending polishing
As for round corners - I can't estimate yet. Haven't digged deep into it. But if it would take few days then I may shift it for one of the next releases.
The 0.58 will come very soon (in development since January). The next other 3 releases will be short (around 1-1.5 months). Probably with round corners.

I think not earlier than in 2020. For Maya too. For example right now round corners are not important compared to other things. In the currently developed 0.58 the GI and mesh lights have much less noise. A lot more robust noise situation for given samples count It will be released very soon. Then in 0.59 there will be node based GUI for Cinema (pretty short update). Then in 0.60 there will massive instancing (tens of millions of elements in scene) and hopefuly support for other particle/scatter plugins. And then all the rest, like more flexible AOVs, masks, cryptomattes, vertex attributes, etc. And then Maya!

Hi xxx man, yes round corner is in the plans, deeper plans. And Maya too! Hope to start doing Maya during 0.6x release versions

Hi Daniel, I know. But there are good reasons for this. Much better noise reduction / per iteration for complex lighting cases. But still need to keep iteration time lower than it was. Pushing this hard!

Hi Marc! And welcome here!
After 2 more releases there will be a documentation explaining how to do many things.
Currently I would suggest to make chrome with 2 reflection channels:
Reflection 1: color = white, roughness = 0.2, IOR = 1.3
Reflection 2: color = white, roughness = 0.6, IOR = 1000 (totally outweights all underlying layers like diffuse and others).
You may optionally add scratched bump for Reflection 2 only. Just make bump and uncheck it for Relction 1.

Yes, Leif! Sorry about that. Mac isn't supported because we are still on CUDA

Hi Chuck and welcome to CentiLeo forum!
Ramp node is needed to create colored output for black/white Input Texture. For example, you may attach some CentiLeo Noise shader to the "Input Texture" slot of Ramp and then create green, blue, red, yellow, etc knots in the Gradient field of Ramp. You may also add Random shader (uniformly distributed greyscale colors) to the Input Texture of Ramp and create colored knots of Gradient to determine how frequent is red, how frequent is blue, etc. in the variety of objects with attached Ramp + Random shader.
You may also add Fallof shader to the Input Texture of Ramp and make colored knots. This way there will be color variations depeding on the view angle.

Kirill

Hi Ikon Art! Welcome to CentiLeo!
There is a very outdated documentation in the download archive, but I think it's not needed.
Actually we will update the documentation after 0.60 release when we will get more speed, node-based GUI, more features (particles, attributes, improved displacement).

There are almost no technical render parameters to affect performance except clamping. Other things are automatic. We try to do it simple and make space only for creativity.

Some valuable info:
1) All the used centileo bitmaps construct caches with the names "*.cntx". At the first bitmap usage it takes some time to construct a cache and also disk space, but later after second launch bitmaps are loaded from caches immediatelly. Later will reduce the *.cntx file sizes. They can be removed manually when the work on scene is not needed anymore.
2) Supported GPUs are with CUDA compute capability 3.0 or more, with 3GB or more. More GPU memory is better of course. 8GB+ VRAM is excellent. Geometry is stored very efficiently, also for displacement, there is a lot of space before you meet out-of-core, which is also very efficient.

Change Log: cntlc4d 0.575 alpha (2019 March 30)
[cntlc4d] Fixed crash at Cinema 4D R20.059 launch

Fixed the issue for R20.059

Change Log: cntlc4d 0.573 alpha (2019 March 22)
[core] Fixed Windows 10 memory bug which could cause crashes in scenes with a lot of geometry and textures.

Thanks Daniel, I will dive into these things when will start converting 0.58 release for Max

Btw, the new CUDA version of NVIDIA that we use no longer supports GPUs of compute capability 2.0 and 2.1. This means that GTX 480, 580, 560ti and other similar devices are no longer supported. See the full list of devices per compute capability here https://developer.nvidia.com/cuda-gpus

Quote
Thomas Jensen wrote: Centileo wether its the bridged r19 or the "legit" r20 release are not working on my r20...0.57 Dont know whats wrong but Cinema is not starting up.

Hi Thomas and welcome here! What are your computer specs?
If this is Windows, Nvidia GPU (with at least 3 GB or even 2 GB) then Cinema may not start if Graphics Drivers are outdated. Please try install new drivers and let me know your results.

Change Log: cntlc4d 0.572 alpha (2019 March 6)
[core] Fixed artifact bug when SSS doesn’t follow motion blur.
[core] Fixed wrong clamping for 16+ bit bitmaps when used for displacement.
[core] Fixed some triplanar mapping artifacts.
[core] Fixed wrong bump mapping over displacement mapping (when bump and displace maps are different).

I have fixed the bug with SSS and motion blur. Will upload today or tomorrow with some other recent bugfixes.

Hi Christopher! And welcome here!

Please, can you upload this file with SSS so I can test it and debug. Just solid light pink and dark pink shouldn't be there?
There shouldn't be anything special with SSS and motion blur. Maybe I have missed inserting time value to some function.
Btw, you can watch how motion blurred object lookd for every time if you change the shutter open/close range (in render settings) from 0..0.5 to the range like 0.3 and 0.3 (just immediate time moment).
There is some limitation for IPR and motion blur: IPR doesn't show more than 4 motion keys in a preview. But if you render from Picture Viewer (or IPR with FIN GEO mode) then you will be able to see all motion blur keys enabled in render settings.

Max ray bounces - limits the total bounces of light path travel in the scene (from camera to light).
Max diffuse bounces - limits bounces from surfaces with non-zero roughness. These 2 settings may be valuable for example when you limit the "Max diffuse bounces" but have a glass object with internal reflections and refractions and for this you increase "Max ray bounces" and get deeper render result and slower render only for this object but not for the rest.

Depth of Field is moving to camera. Focus distance is controlled from camera object. The rest things (like sensor width, blades) will land there too once I dig there again.

Caustics are currently blurred. However what you see around semi-transparent objects is not caustics but a fake transmission with forced IOR = 1 (for faster render in general case). Real precise caustics and dispercion will be addressed later when other production features will be implemented.

Metropolis is currently not used internally. But it was in the past. Even after all improvements that we did for it there are still so many problems for animation, motion blur, slower for everyday production scenes, many ray bounces and etc. We will continue to investigate though.
Anyway, the next 0.58 is about render time optimizations, so stay tuned