Ideas on making V-Ray's sampling UI more intuitive.

[RockinAkin]

I was able to give the new 'Divide Shading Samples' checkbox a test drive yesterday, and I've gotta say that I'm very thankful to now have the option to control my sampling using either method.

I'm finding it effective to begin by setting all my local secondary subdivs to 1 across the board, tune the AA subdivs to what's needed to properly resolve geo/textures/dof/moblur, and then systematically increase my local secondary subdivs as needed to eliminate noise in those specific aspects while keeping render times at a minimum. I definitely feel more in control of my scene, and like I said - it definitely feels good to have the option available.

I've added its functionality to the DMC Calculator as well for anyone interested.

[jstrob]

I recently tried to understand Grant, Akin, John and Toni explanations about the dmc sampler and the first thing I wanted to do after understanding them is to make this exact request akin Just made here! But I would add a few features to the interface:

I would create a different global multiplier for each kind of noise source. and I would rename the software as "VRay Intuitive 4.00.00"! And the feedback boxes in the material and sampling areas are a must or we need to have akin claculator open at all time when using vray... Is there anyone who could make a script that open a window and shows something like akin's calculator witht the vray settings taken from the scene and showing the resulting samples?

That way we can balance the sampling for every area and if we need more precise balancing then we go right in the mat, light etc themselves.
And the global mutliplier for the reflection and refraction could be linked to the glossiness value, lower glossines would get proportionally more subdiv.


And Vlado I don't understand your question about where would the irradiance map and importance sampling fit into this? First can you tell us where do they fit right now? I don't use irradiance map so I don't know about it. And importance sampling is something vray does under the hood and I can't find any info about that other than vray check if it's bright or dark and put less sample in the dark area. So I guess it would just continue to work that way? Or is there a settings that influence importance sampling that we are not aware about right now?

Also about your comments saying human time is more important than machine time... I would say that in my case and the case of everyone working with a deadline, our machine time really has an influence on money income and money = human time. I mean if I miss a deadline or don't meet quality requirements within my deadline I can loose a client and a lot of money meaning a huge human time loss...

[vlado]

There already are different multipliers for the different sources of noise - for each of them in fact - be it material, light etc. I'm not convinced that adding yet more parameters is a good thing that will bring any particular usefulness other than more confusion. After all, it wouldn't really do anything more than what is possible right now. Ideally, what we should be striving towards is the exact opposite - removing any parameters that bring confusion and trying to make V-Ray as automatic as possible.

Best regards,
Vlado

[jstrob]

There already are different multipliers for the different sources of noise - for each of them in fact - be it material, light etc. I'm not convinced that adding yet more parameters is a good thing that will bring any particular usefulness other than more confusion. After all, it wouldn't really do anything more than what is possible right now. Ideally, what we should be striving towards is the exact opposite - removing any parameters that bring confusion and trying to make V-Ray as automatic as possible.

Best regards,
Vlado

Yes but they are not global. That's why we need to create scripts to do that. I mean if I have a scene with 200 light and 78 material, I can't change them one by one or if I work on someone else's scene I don't want to have to go into each light and material to see how they are set. With a single place for all the global multipliers we see at once how the scene is set or at least we have a better global idea.

[jstrob]

And Akin is right about the counterintuitiveness of a setting that doesn't do anything to the end result when changed. I think that if everyone could just see what's happening to the 4 sampling values with the feedback boxes Akin put in its interface images (min max image sample and min max dmc sample), VRay would be more intuitive.

[vlado]

Yes but they are not global. That's why we need to create scripts to do that. I mean if I have a scene with 200 light and 78 material, I can't change them one by one or if I work on someone else's scene I don't want to have to go into each light and material to see how they are set.

This particular problem can be solved in other ways, f.e. by providing a tool to give you all these settings listed in one place, and potentially giving you the ability to mass-edit those settings, similar to the light lister.

With a single place for all the global multipliers we see at once how the scene is set or at least we have a better global idea.

No, you just have yet one more place to have to check.

Best regards,
Vlado

[jstrob]

This particular problem can be solved in other ways, f.e. by providing a tool to give you all these settings listed in one place, and potentially giving you the ability to mass-edit those settings, similar to the light lister.

No, you just have yet one more place to have to check.

Best regards,
Vlado

Yes I understand. I think you are right on both points. And a tool like this would be awesome.

Maybe something like the Akin calculator could also be in a tool just to better see what happen under the hood.

[RockinAkin]

Yes I understand. I think you are right on both points. And a tool like this would be awesome.
Maybe something like the Akin calculator could also be in a tool just to better see what happen under the hood.

John Martini has been working with me to script a tool to do exactly that.

It will grab all the subdiv values throughout your entire scene, list them one place, and show you the resulting min/max samples that will be taken (using the formulas from my DMC calculator) in the context of your scene's setting.

Hopefully it'll be ready for testing soon.

[jstrob]

John Martini has been working with me to script a tool to do exactly that.

It will grab all the subdiv values throughout your entire scene, list them one place, and show you the resulting min/max samples that will be taken (using the formulas from my DMC calculator) in the context of your scene's setting.

Hopefully it'll be ready for testing soon.

Ok wow! That's great. If you need testers don't hesitate to communicate with me. I was considering trying to script something myself but I'm not a master scripter.

[LudvikKoutny]

It's really ridiculous how much math needs to be done in order to do something as essential as setting up scene shading and lighting

[Art48]

Noone said 3d stuff is easy. Although getting more user-friendly should always be the goal.

[vlado]

It's really ridiculous how much math needs to be done in order to do something as essential as setting up scene shading and lighting

You don't NEED to do it!!!! Use progressive rendering, leave it for as long as needed and you are done. If we removed all subdivs values from V-Ray (and I must say that I'm very very tempted to so this), this whole problem would not exist... It still mystifies me why people want to make their lives so complicated... they don't care about how many samples any other render engine makes, it beats me why it is such big deal here.

Best regards,
Vlado

[jstrob]

You don't NEED to do it!!!! Use progressive rendering, leave it for as long as needed and you are done. If we removed all subdivs values from V-Ray (and I must say that I'm very very tempted to so this), this whole problem would not exist... It still mystifies me why people want to make their lives so complicated... they don't care about how many samples any other render engine makes, it beats me why it is such big deal here.

Best regards,
Vlado

I don't understand Recon442 either, we are not talking about doing math. But just about a way to understand how vray balances the sampling between geometry silhouette, motion blur, dof on one side and GI, shadow, ref and refrac glossiness and SSS on the other side while using 2 tools: primary ray samples and secondary ray samples. the math is thankfully done by the software, all we do is turning the knobs to make adjsutements.

Then why some people use Arnold or Maxwell and don't care about the number of samples? It's because they work on multi hundreds millions budget full feature films and their deadline is in 4 or 5 years and they have access to thousands nodes render farms. Or maybe just working on personal projects without any deadlines...

But when we work on more limited budgets and tighter deadlines we need to render as fast as possible with limited amount of render nodes. That's when we need VRay and we need to know exactly where the machine time is used for and for exactly how long depending on which sampler is doing what. And that's when we need such tools as the one we are talking about here.

[joconnell]

Then why some people use Arnold or Maxwell and don't care about the number of samples? It's because they work on multi hundreds millions budget full feature films and their deadline is in 4 or 5 years and they have access to thousands nodes render farms. Or maybe just working on personal projects without any deadlines...

Yep they work on huge budget films with pretty big render farms but they also get massive render times - a friend is using arnold and vray for commercials and he hasn't been able to render an ad at 1080p in arnold, only 720p which is okay for our local market. In larger studios I've heard of 100 hour per frame render times so even if you have 100 nodes for a 100 frame shot, you're still going to be waiting 4 days to see the results.

People that use arnold and maxwell don't complain about the samples as they have no choice - they have no control to make their render times faster so they just accept it and move on, as a result if you want to use either render in cg heavy productions, you've gotta pay for the render power to support it.

[LudvikKoutny]

You don't NEED to do it!!!! Use progressive rendering, leave it for as long as needed and you are done. If we removed all subdivs values from V-Ray (and I must say that I'm very very tempted to so this), this whole problem would not exist... It still mystifies me why people want to make their lives so complicated... they don't care about how many samples any other render engine makes, it beats me why it is such big deal here.

Best regards,
Vlado

I never touched for example settings in DMC sampler past year. But there is still that inconvenience of adjusting material and lighting sampling, as well as balancing AA versus secondary sampling, which is now divided to two parameters for even more inconvenience. Basically you got the thing that divides shading subdivs, so increasing max. adaptive AA subdivs not only increases AA, but also reduces secondary rays, and then you got also min. shading rate, that somehow biases this entire thing, by clamping the minimal amount of secondary rays from bottom up. Sure we can disable shading subdivs division, but i still stand behind my opinion, that the checkbox is broken, because even when i disable it, there are still situations where i can achieve very different noise level at exact same rendertime, or same noise with very varying rendertimes depending on how well i balance out the sampling.

That being said, i do not ever disable shading subdivs division. Non the less, that universal approach you are suggesting simply does not work well. Sure it renders scene clean at some point (in distant future), but sometimes rendertimes are just ridiculous. I tried to apply this universal approach on several past jobs, but i always ended up using my setup, as i mostly got a lot better speed/quality ratio. Sometimes even twice as good. So yeah, i have to do it, if i do not want to waste my rendering power. If you want, i can even send (Or at very least try to convince my boss to allow me to send) you the scene where universal approach simply fails.

Some of the other renderers have a simple benefit of having less settings, but still rendering with somewhat optimal speed/quality ratio.

Lastly, my post wasn't implying i am doing lots of math to set up my scene. I don't... really For me, it's quite annoying, but still not very long process. There was a bit of sarcasm to my post, as in my opinion, having some calculator that shows some exact sums of sampling values isn't going to help anyone, as noise levels under given sampling settings are varying across scenes. But non the less, one really needs to have some sort of feel for the sampling balance, and some skill of prediction to be successful with Vray. At least if optimal rendertimes and quality are desired. Otherwise, sure... you can just let cook images till infinity. That's how NextLimit decided to market Maxwell