Returning to the octane

[Gragh_Sparrow (Ilya Denisov)]

Actually, at first I wanted to share just two renderings of the exterior, but in the end I decided to share my thoughts on the octane and cycles.

After a year of working with cycles, I’m glad to be back to octane.

After a number of sleepless nights trying to understand something, I came to OCIO Agx Punchy and using EXR to get some details in post processing.

Octane is still several heads higher in terms of usability than cycles, it is incomparably easier to achieve realism in the picture, or at least the WANTED result.
In some cases, Cycles just doesn’t give you the color you want, so you have to artificially turn it up or down, add or subtract saturation.
A big plus of Cycles is the ready assemblies from the addons. And also responsiveness and quickness - they’re better than in octane. And denoise - on large areas - walls and ceilings, floors - it works very well, much better than in octane. Generally speaking, out of the cyls I usually used full denoise and sharpness from k-cycles, and then more photoshop, in octane I take about .8 denoise, depending on the situation. But I mix in photoshop, it’s more convenient. The only downside, perhaps someone knows - I have problems in octane with denoise when viewed from the camera. It works fine in viewport, but for some reason it turns on when it wants to, and very rarely in camera view.

I find it difficult to summarize their actual “speed” of rendering, and work - in octane I achieve other results with a different amount of force. In a number of renders I’m happy with Cycles as well, but the result is slightly different. And it is different visually, how much force it takes to achieve in it what I achieve in octane, it is difficult to say. The issue is that it all adds up and not otherwise, each rendering pushes a certain result, and to achieve what you get by itself in the other - you already need additional adjustments.
As for speed - because of a different denoise, and I would say a different “noise”, I can not say what is faster and what is slower. Yes, I can get the final picture in cycles faster, the final picture in the limits of what will be accepted, or “eaten” by the customer, because they are often omnivorous. But it will not be the same quality as in octane, in the matter of details and accuracy, at the same time octane will take more time to clean the picture. The ones I enclose are not yet final renders.
In general it seems to me that Cycles can produce a picture on lower samples, but even with increasing samples it is difficult for it to improve the quality, in turn Octane requires more samples, time, but the result will be better.

Yes, it’s different styles, different lighting solutions, but in the end it’s a different way and a different system.
Have a good day.

Octane

Cycles

[bartv (Bart Veldhuizen)]

I featured you on BlenderNation, have a great weekend!

Best of Blender Artists: 2022-48

[BD3D (Dorian B.)]

Very nice work!

[alekba (Alek)]

Nice renders. My personal favorite is Cycles exterior, not because is better than others… just personal taste. Love to see Otane version, for comparison, side by side.
Because you have right, Cycles and Octane give a different result. Octane is more like Corona, sharp, crisp, precise. This don’t necessary mean that Octane is better, especially not in every case. But many people prefer this look, especially in archviz.
Strange choice of short focal length for interior shoots. Or I’m used to seeing a longer focal length

One question. I hear about OCIO AGX, to this work with Octane, and can you give me some links with more information’s.

[Gragh_Sparrow (Ilya Denisov)]

Hi, https://www.elsksa.me/scientia/cgi-offline-rendering/rendering-transform

Regarding sharpness, I’d say it’s described very simply - in Cycles I rarely wait longer than 5 minutes for a render, with denoise and the render itself allowing for an acceptable picture. But not because it’s detailed or cool, but because it allows you to get a “cool” picture in blurred detail. For a couple of reasons - it’s harder to wait for a cleaner picture, and that if the picture is very cool calculated, it will show that there is a lack of detail, plastic. And the more details you add, the less again will be calculated). It’s a vicious circle. And then there’s the faking of the roughness in the Cycles denoise.

In octane, it really is a slightly better calculated picture, to which I then apply sharpening in Photoshop. In Cycles too, but it’s less visible, maybe. It’s the same with Corona - it just makes a much, much less noisy picture and the sharpness in it is added in the render window itself - so we often see much less sharp, and generally lower quality work done in blender than the overall selection in Corona.

[alekba (Alek)]

Sharpness which someone does in post is different from sharp - crisp vanilla render. It’s not same. Of, course I talk about my personal experience. For me this look differently.
Noise, annoying “inconvenience” and denoiser, detail killer
I have impression that little dose of noise is “allowed” these days. Somehow this overlap with GPU renders appearance. People mix noise which they add in post to camouflage noise - artifacts. Year or two ago read article about this topic. Couple good archviz dudes talk about his technic.
Heh, couple years ago level of noise in his renders was “inadmissible”.

Thanks for link.

[Julius_Moorman]

Stunning work man!

[Gragh_Sparrow (Ilya Denisov)]

I agree about the noise, I even described in the comment of one of my works, what and why I add it in a number of works.

But I would also say that there is no magic in the really fundamental question of rendering in Corona. I would say that first of all they render very well the details, like gaps between boards, recesses in cabinets, details in bevels - I think it’s just thanks to the cache, which allows you to remove noise in a way that you would not be able to calculate for yourself in Cycles or Octane. These details become visible in the first passes, simply because of the cache. And further, the picture is counted to very clean very quickly. In the case of cycles, octane - it is almost impossible to get rid of noise. Also - what filter size is used there? In octane the default seems to be 1.2. And in the end I think even the distribution of noise is very different - the edges of objects, slits, but globally on planes noise level can also go away not immediately, so it turns out that the sharpening is higher. It is a question of time and drawing technique. I would not say GPU, because there is Redshift, I never used it, but if I’m not mistaken, its “sharpness” will be the same as in the Corona and due to the same principles.

And here given all the nuances above, it turns out that yes, we can achieve identical or almost identical results, moreover, in quality even higher in octane, but the question is that it will be one case for which we will not be limited in time. And when it’s a job, you won’t spend extra time on the project and it will turn out that 10 out of 10 in octane will get a typical picture for octane, in cylinder for cylinder, and in corona - for corona.

Lux, Corona, Cycles

I tried to compare different engines in order to understand for myself what the advantages of each of them would be, and what are the disadvantages. I made a comparison in Corona, Cycles and Luxcore. To be honest, I’ll bring the sequence, so I started to make a scene in the Lux to try it out, then I decided to make a comparison with Corona, and in the end I decided to do it in Cycles. Accordingly, the process was such that in Lux I relatively easily tuned materials, light, and I ended up with, I…

And are they so different fundamentally? It’s a matter of nuances, but which create the overall result.

Again, you noted that there is a “difference between sharpness” but, what is it? If we take a picture from Octane or Cycles and add sharpness, a little bit, with the right parameters, in photoshop, not overdone, but correct. How will the picture be different from the Corona?
This is an image, not a vector, there are no fundamental differences, technical, in the way sharpness is obtained. Maybe the developers of Corona just made the stock result so, maybe they have higher sharpness.

Otherwise, it’s a question of noise. But yes, I think so too, that it is there and it is visible.

And it’s not an issue of added noise in photoshop or in K-Cycles, no, it’s an issue of the fact that here we have plaster on the wall and in the Cycles without denoise with pass albedo and what else is there, normals? Here without denoise we’ll never see that and it’s not a question of artifacts, it’s a question of being able to see that result in the Corona. Without any tricks, without “adding” noise, we will see it in one place, and in another place only because of the extra passes in the denoise.
Yeah, I just remembered why it became possible to do interiors in 5-7 minutes in Cycles - because denoise has albedo and normal passes, without them it would be as unrealistic as it was 4 years ago. Even with my 3090. And here’s the payoff for that, fake(or clumped?) roughness on smooth surfaces. On such materials, it’s easier to achieve visibility through colors, because we’ll barely see the roughness.

Maybe if you put renders for an hour in Cycles, the result would be better) then correct me. In general, I will be glad to hear other people’s opinions.

[alekba (Alek)]

Redshif is not on sharp side. He preserves details… so, so… But I think that Octane is better regards details.
Redshift is on soft - warm side. Here is one great example. Rare example where you can compare result from some good artist, which know in depth both renderers.
Same author same scene, render in Octane and Redshift in C4D. Software don’t matter in this case, because renderers “must” behave same… I guess.
ZAKI ’ on Behance
ZAKI ’ on Behance

Couple years ago, I see comparison between Vray and Faststorm in 3D max. Faststorm is like Octane, Corona, sharp and precise. Vray is more warmish, soft. Dude which posts these renders are very good in archviz. btw. he compare speed with his dual Xeon workstation ( couple thousand $$$ ) and single GTX 1080, which was in this time top GPU card. Good old days, when you don’t need to sell kidney, leg to buy new Nvidia toy
However, good artist can achieve excellent result in every modern renderer. Except in particular cases all are appropriate. Rest is on personal taste, and of course his “majesty” customer

[Gragh_Sparrow (Ilya Denisov)]

Compare… two different scenes? What was there to see?

Somehow you’re relying on subjective criteria like softness and… warm?

Here’s a definite example. And it’s not the interior, where it’s almost impossible to get rid of some noise in an unbiased rendering.

And that’s probably on the brute force.

Fstorm

[alekba (Alek)]

I give examples of same COMPLEX scene, with complex materials, volumetric, etc. Models are same, procedural materials look very similar, only camera angle is not matched. So how this are different scenes?
Examples with simple subject like this bust are like sphere on plane example.

However as far I know you can force Redshift to be unbiased like, if chose brute force as second engine.
Of course, this will make, much slower. But I don’t.t think that this will change look of Redshift.

Many years ago, I couple times use Vray without GI, just for fun. After “downgrade” I put zillion lights around to fake GI, just like in old days. This don’t change overall look of renders.

However, little amount of noise is not big NO, NO like couple years ago. And with denoiser which become better and better this will be even less issue in future.