OTOY Forums

caustic soda is great, but it is not always necessary. it would be good if it is possible to enable or disable this feature.

Not sure what "caustics soda" is, but i think i know what you are talking about. The answer is no. Caustics are a natural phenomena and since octane is using physically correct light simulation caustics happen where they should. So there is no on off switch like other render engines. You can take measures to reduce them such as no small light sources pointing through specular materials but i am not sure why you would want them off. It does not increase render time because like other engines because its not a secondary calculation.

Hope that makes sense.

Take a look at this request almost 1 year ago: http://www.refractivesoftware.com/forum ... f=9&t=1825

Even self proclaimed for the most "physically proper unbiased" renderer around - Maxwell Render, has this future - switching indirect/direct caustic calculations on and off for refractions and reflections along with illumination switch - indirect and direct..
Imagine how this will lower the memory footprint in some cases.
This will lower render times pretty much as well and it's nor required in most cases..
In this case - Octane Render, I think it's a must, even if the MLT core will speed things up in this area...

andrian wrote:Take a look at this request almost 1 year ago: http://www.refractivesoftware.com/forum ... f=9&t=1825

Even self proclaimed for the most "physically proper unbiased" renderer around - Maxwell Render, has this future - switching indirect/direct caustic calculations on and off for refractions and reflections along with illumination switch - indirect and direct..
Imagine how this will lower the memory footprint in some cases.

maxwell.gif

This will lower render times pretty much as well and it's nor required in most cases..
In this case - Octane Render, I think it's a must, even if the MLT core will speed things up in this area...

I have been thinking about adding a new animation-centric render kernel,
or upgrading directlighting to it.
Basically you would have full manual control, eg direct lighting, with optional ambient occlusion, and depths per BRDF type (eg you can configure X amount of diffuse, X amount of glossy and X amount of specular bounces, with options for reflection and transmission), which due to configuration can also cut away caustics out of the rendering equation.

I developed a kernel like this which was rather unique for luxrender back in the days, to use it for animation rendering for a french movie.
I still think it's a viable option, for a select number of applications.
Eg, those wanting absolute speed with a tradeoff for quality will be able, given more complex configuration (like vray), decent rendertimes for projects that don't require absolute unbiased/path traced quality, eg animations.

You could for example, configure it with maxdepths for specular reflection and transmission (separately), and then add 2-3 bounces of indirect diffuse (diffuse <> diffuse only, eg no caustics or diffuse <> glossy), and add 1-2 bounces of glossy reflection.
This will give you a decent looking image for animations, compared to a pathtraced image it will lack some 'depth', but will still be very nice and acceptable for animations, with no fireflies, and very fast rendertimes. (arnold does something similar)
The problem with this approach is that it's complex to configure (although presets might help), and it will be very difficult to use for scenes where unbiased renderers really shine at currently (complex architectural interior renderings for example)
Kernels like this are perfect for outdoor scenes, and propperly lit (with several area lights, unoccluded) animations of subjects, eg cars, characters, products, etc...
The problem is that they do not offer any empirical / configuration-less quality, and they need to be finetuned for the scene in question with great precision. They also are completely unusable for some scenes (as stated above, complex interiors, and light beam/trap experiments etc...)

We would like to focus on the new unbiased render kernel with importance sampling and efficient caustic rendering first though, i will keep developing an improved concept of a 'kitchensick' renderkernel in my head and we will see what happens after 2.5 timeline wise. (we need to focus on finishing and releasing v1.0 final, that's our main priority)


Radiance

Radiance:

Thanks for your thoughts.

On one hand, there are users that will enjoy having a lot of different render kernels that can do different things. But on the other hand, there are other users that want simplicity and will find having lots of kernels to be very confusing.

It would be great if you could just create one kernel that could do it all (with options for various things that could be turned on and off), but that is probably not possible.

Radiance, that sounds great!

I don't care how much kernels or settings are there - the more, the merrier. Actually, having separate kernels with clearly defined roles is much better than one with all the options crammed together. Even total noobs will be able to understand that if you are doing animation you use X, if you are going for realism you will use Y. It's not possible to dumb it down further

I really need the improved direct lighting kernel
I am working usually in tight deadlines and a nice ambient occlusion solution that supports transparent objects and other adjustable features would be really great.
Please go for it. It is worth not only for animations.
Pathtracing a scene over a couple of hours is nice but having a fast "fake solution" is although a reason why so many like octane
Thank you.

Octane first attracted me with its speed. Then he drew the physically correct illumination allows to make photorealistic images without a million options.
It's great, a lot of engines for different tasks. It's not confuse the user interface, but only make it flexible.

yes, I fully support a new kernel for animation, flexibility is important.

I couldn´t agree more for a direct lighting animation kernel - I´d even buy a next development stage to prefund it, or another license if it would be a seperate product.

Currently there isn´t really a fast "semi-biased" or "semi-unbiased" Renderer for animations with a nice pricetag. For the most part I actually use Blender internal, if you know your way around (well, like in every raytracer) you can fake quite a lot no one notices and it works with decent speeds. For some reason nowadays everyone absolutely needs to have a 101% phyisical correct raytracer and it seems that we don´t even need control over certain features, because what do artists know. It seems like rationality has to win over passion and technicalities over artistry.
For many, many render applications it would suffice if it looks good, rather than correct (which many can´t spot anyways, especially in animations)

I mostly do image movies, trade show presentations, product visualizations for construction plants and stuff like that. So it is either a rather "technical" style where I do not need photorealism but control over certain features to speed things up, where I use Blender Internal, or it is photorealistic, where I use Octane - which is kind of uncomfy for animations but gives great results in no time.

I think offering a direct lighting animation kernel with a, let´s oversimplify:
"reality -------[50]------- speed" slider and not so physical correct features would place Octane also in a part of rendering where there isn´t much competition at the moment - at last none that I am really aware of, but please, don´t hesitate to rub my nose in the tool I am looking for if it already exists.