Optimal render settings for Hair?

[Rectro33]

Hello.

I own MODO, and C4D, but have Vray for C4D at the moment. I understand that Vray is basicly the same across all platforms, and unfortunatly the Vray4C4D forum just is not giving me the support I need, so please bare with me.

Iv reverting to looking at Max, Maya manual as there is non manual for C4D beta v3.3. Im trying to get the best settings for hair, but im getting nowhere. BF.LC is the only thing I know iv got to have set. Other settings that I see being mentioned must be named different as I cant translate what im finding on the net.

I need to know what is the main settings that are responsable for noise, and shapness. I know AA or as its called image sample is for sharpness. Min shading rate of 1 should in theory give me the best render. Sub div min1 max 100, noise 0.005. Can you see anything wrong here?

DMC Sampler, bit sure what this is, and if its even being used, but its set to Adaptive amount 0.85 Min samples 16, noise freshhold 0.005, gllobal sub div multiplyer 1. Iv put light samples to 24, and for the hair object iv got the compositing tag with Mis settings, secondary Rays Multi 0.3

The results im getting is noise, lost of it.

If there is any videos out there on rendering in Vray that will help demystify the settings by example, then please point me in the right direction. At the moment it feels like im totaly alone in figuring Vray out, and am starting to regret that I purchased it.

Thanks Daniel.

[vladimir.nedev]

If there is a facility for exporting .vrscene files from V-Ray for C4D, you can export one and send it to me to take a look at.
E-mail is vladimir.nedev [at] chaosgroup.com

Greetings,
Vladimir Nedev

[^Lele^]

It's very early days, and we couldn't possibly expect to cover all possible corners of scene setup, so I hear you on feeling like being alone, Daniel: you ARE among the first ones!
There is, however, a wealth of information for the Max and Maya versions, most of which will translate with very little effort in the Modo workflow.
Not pretending to debug your scene, one rule of thumb is to not touch minimum shading rate at all, from its default of 6, as that will prove less efficient more often than not.
V-Ray automatically reduces it for the areas which would benefit the most from it (f.e. hair primitives), while insuring it stays near optimal everywhere else.
MSR will eventually disappear.

Also, don't touch any of the DMC sampler settings, from their defaults. Those too will soon disappear, as the new sampler introduced for 3.3 (for Max and Maya) renders all that tweaking obsolete and counter-productive.

This linked cheat-sheet (and related thread) shows the preferred way, bar none, to go about a clean image with V-Ray, whatever the app it runs in, after 3.3.

http://forums.chaosgroup.com/showthr...image-with-SP3

[Rectro33]

Thanks for your reply.

Im not using Vray in Modo, I own MODO but am using Vray in C4D. I posted here as the offical forum for C4D is not Chaosgroup, but is run by primarily a very helpful Architect from my understanding, the user base seems to be 95% or more toward architecture which means I am in a very small minority who wants to use Vray for portraits, character work hence limited knowledge, and interest in my area, this is what leaves me feeling alone in learning Vray.

If I had Max Maya, and got Vray for that then id have tons of support and threads to go through from thoes who have been using Vray for years, and have found methods which save time in the long cave man style trial and error im having to go through.

I wish Max, was transferable, but its not as easy as you may think. The layout is so different, some options are either not there in C4D version, or are setup in a nodal workflow. It also seems there is plenty of great examples of portrait work, but not any in depth training on this subject, maybe a few threads showing some w.i.p info, or generalised videos showing a overview. Im thinking this is the trade off when moving to the big pro render engines, its assuming all its users are professionally trained, and have a good education in photography, and complex render engines. I dont think Vray is half as complex as I imagined it to be, its just getting a grasp on what does what and what visually it produces. Im not lazy when it comes to learning, but its the time factor I feel creeping up on me that has me doing more experiments that actual art work.

Dan

[vladimir.nedev]

V-Ray automatically reduces it for the areas which would benefit the most from it (f.e. hair primitives), while insuring it stays near optimal everywhere else.

Lele, I think this is true only for V-Ray for 3ds Max. I don't think it is automatic anywhere else.

The results im getting is noise, lost of it.

I did get the .vrscene to render with less noise, but the render times increased as well.
Is it ok if I post some renders from your scene ?

Greetings,
Vladimir Nedev

[Rectro33]

Thanks for your time to look at the scene, please do show the results.

Iv got a bit better renders since I sent you the scene, but only since iv upgraded my computer from a i7 2600k processor, to a 5930k with double the ram at 32gb, before the upgrade the render times where astronomical, and always seem to stop before the noised cleared.

Thanks, Dan

[Rectro30]

These took 18min to render, but its my computer thats holding its own in this case, iv not found optimised settings yet. There is post work done on these first two. The 3rd image is raw image.


Dan

[vladimir.nedev]

First a few things I noticed about the .vrscene:

- it uses the progressive image sampler, but the new adaptive method in V-Ray 3.3 (VBAS - variance-based adaptive sampler) is not enabled.
There must be a SettingsImageSampler { dmc_adaptive_method=1; } line for it to be enabled.
This should be reported to the V-Ray for C4D developers. It is not obvious by the name, but the parameter affects both the Adaptive and the Progressive samplers.

- the 0.3 subdivs multiplier is not applied to the hair geometry, only to some V-Ray proxies, as far as I can tell. Not sure if this is a problem with the V-Ray for C4D exporter or the parameter was set incorrectly in the scene.
This means that the hair will use the normal shading rate, which seems relatively fine in this scene.

- noise threshold is 0.016

- the color mapping mode is set to Color mapping and gamma, this means that the color mapper will apply both the dark/bright multipliers and the gamma (which is 2.2 in this case).
In this case, the sRGB checkbox in the VFB must be disabled, as otherwise the gamma is applied twice.
If the gamma is applied from both the color mapper and the VFB, V-Ray will see a much darker image while looking for noise, compared to what the user sees.
This results in much more noise than expected.

- Irr Map / Light Cache is used, which seems fine, I did tests with Brute Force / Light Cache and they did render slower

- the "simplify for GI" option of the hair material is disabled

Here are my results, on a i7-2600 CPU 3.4-3.8GHz.
The frame stamp I copied from V-Ray for MODO and forgot to remove that part, but the renders are done using V-Ray Standalone.

1. The render before any changes to the scene.
Render time : 3m 29s


2. Added the dmc_adaptive_method=1; line, so the new V-Ray 3.3 adaptive method is used.
Render time increased significantly, but there is significantly less noise.
I think this is one of the characteristics of the new adaptive method.
It samples bright and dark areas more evenly, whereas the old one under-sampled dark areas and over-sampled bright areas.
Since the image is pretty dark, the old adaptive sampler renders fast but results in a noisy image.
Render time : 8m 38s


3. The albedo of the head material is 0.8, which is quite a lot for human skin.
According to google, the maximum albedo of human skin is around 0.6, so I set it to 0.55 and did a new render.
This reduced the render time slightly as less light is bouncing around.
Render time : 8m 12.8s


4. The hair material settings look a bit strange compared to the presets in V-Ray for Maya.
The presets have a very low primary specular amount (0.25), while it is very high in the .vrscene (1.0).
The presets have a lot of transmission usually, while it is very low in the .vrscene.
The presets have the "simplify for gi" option enabled.
First I tried the "white shiny" preset.
Render time : 6m 35s


5. Then the "blonde shiny" preset.
Render time : 6m 27s

[vladimir.nedev]

6 Finally the "red shiny" preset.
Render time: 3m 37s


Hope this information helps you.

You should report to the V-Ray for C4D developers that they need to set the dmc_adaptive_method parameter for the Progressive sampler as well.
Otherwise the new adaptive method is not being used.

Greetings,
Vladimir Nedev

[Rectro33]

Vladimir, thank you so much, this is fantastic support, I will go through all this in detail tomorrow. 6min for a render is a dam site faster than 18min with a 5920x clocked at 4.5ghz, even 8min is still great! The 0.3 subdivs multiplier is applied to a comp tag in C4D under Misc settings, so maybe it didn't carry over on the export as tags are a unique thing to C4D.

If you render my scene without changing anything what render times are you getting?

Thanks Again.

[vladimir.nedev]

The 0.3 subdivs multiplier is applied to a comp tag in C4D under Misc settings, so maybe it didn't carry over on the export as tags are a unique thing to C4D.

It should carry over. Even if you render from inside C4D, the V-Ray for C4D plug-in still needs to create an internal representation of the C4D scene as V-Ray plug-ins, that is pretty much the same as the .vrscene.

If you render my scene without changing anything what render times are you getting?

The first render was without changes. It was 3m 29s and the noise threshold of 0.016 was reached. Note that I am using the V-Ray 3.4 Standalone that comes with V-Ray for MODO.
I can't remember any notable speed improvements to hair geometry or hair material compared to V-Ray 3.3.

Greetings,
Vladimir Nedev

[Rectro30]

Hmm, so thing is very wrong? How is the hair being produced, I mean it's from c4d so assume when the hair is made its not polygon data, I know natively modo hair is not the same as c4d hair, so vray is converting hair into its own method which transcends across all vray implementation meaning modo, Maya, max, C4D when using vray is all the same? . What's making the scene iv sent you render on 3min, when mine is taking 18min. I think I even had to set min render time to 15min in the scene? I will upload a screen shot today.

Thanks, Dan

[Rectro30]

Hi.


According to the scene I sent you, I had RT Turned on, this had Max render time set to 12 min which was literal sampling on and on with diminishing returns, how are you able to get variable render times if its set to 12? I had assumed that Max render time was a limit rather than a target hence why my render times where so long. I turned off RT set the Progressive Image sampler Max time to 0, now my render times are 1.38 I can only assume at the time I set this up that I had concluded that even at 12-18 min there was still noise it needed even more time, but this is not the case, I need to know what settings will give me less noise rather than let RT engine render on and on making little difference. The noise id like to clear is realy at the bottom of the hair where it becomes finer.

Simplify for GI in the hair settings had taken render times from 1.38 to 1.27, but there was a noticeable difference in quality that being losing contrast, nothing that level Post cant fix.

In this picture your see that in the Compositing tag iv set the multiplayer to 0.3. I put this setting up to the multiplayer of 1, and its made zero difference in the render time so can assume either this is broke, or iv done something wrong?

Thanks again for your help, I feel im finally getting somewhere now.

[vladimir.nedev]

Hmm, so thing is very wrong? How is the hair being produced, I mean it's from c4d so assume when the hair is made its not polygon data, I know natively modo hair is not the same as c4d hair, so vray is converting hair into its own method which transcends across all vray implementation meaning modo, Maya, max, C4D when using vray is all the same? .

It is represented internally as the same primitive (i.e. tubes/cylinders), whether it is coming from Max/Maya/MODO/C4D, so this shouldn't have an effect.

According to the scene I sent you, I had RT Turned on, this had Max render time set to 12 min which was literal sampling on and on with diminishing returns, how are you able to get variable render times if its set to 12? I had assumed that Max render time was a limit rather than a target hence why my render times where so long.

The Progressive image sampler and RT have separate settings for:
- max render time
- max paths per pixel / subdivs
- max noise
When one of these is reached, the rendering is stopped, so yes, all of them are a limit, not a target.
When a scene is rendered with V-Ray Standalone, it uses Production rendering by default, in this case with the Progressive image sampler (there is a command line option to make it use the RT CPU/GPU modes).
I guess my rendering completed so fast because it reached the noise level (0.016) of the Progressive sampler.

The noise id like to clear is realy at the bottom of the hair where it becomes finer.

This is AA noise. Since it seems that you can't set the secondary rays multiplier for hair, you may have to decrease the min shading rate to something like 2.
But this may increase the render time for other areas of the image where there is no hair.
You can also try making the hair a bit thicker, but I am not sure how this will affect the final look of it, and whether it will be acceptable.

In this picture your see that in the Compositing tag iv set the multiplayer to 0.3. I put this setting up to the multiplayer of 1, and its made zero difference in the render time so can assume either this is broke, or iv done something wrong?

You can try some of the other settings that have a more obvious effect in order to confirm this.
For example, disable the hair from primary visibility in the "Comp Properties" tab, or make it a Matte object from the "Matte Properties".
If this too doesn't work, then we will be sure that the tag doesn't work on hair objects and you can report it as a bug to the V-Ray for C4D devs.

Greetings,
Vladimir Nedev

[Rectro30]

You are right, the comp tag is not working at all on Hair objects.

Thanks, Dan