OctaneRender V4 XB3 Tweet

[Notiusweb]

Could you please post a screen shot of a simple Material Map set-up, doing whatever it does now?
I can't figure out link-wise how it plays into a mix of FBXs, say for example.
I am imagining, as I have 2FBXs, with equal # material node channels, and one FBX with mats already set up, another without any, and this thing (Material Map Node) could link the two to transport the maps from one to the other (?).
But, even though I am a master of nodes, and all nodian things, I have node idea what to do with the Material Map.
Do you node what I am saying?

THX!

PS- Say Node to Drugs!
(Except on weekends.)

[Hesekiel2517]

Did anyone get AI Light working? I tried it on multiple scenes. The results are always worse (more noise/longer rendertimes) than without it.

[haze]

Did anyone get AI Light working? I tried it on multiple scenes. The results are always worse (more noise/longer rendertimes) than without it.

Wallace has made some major improvements in this area for XB4, a lot of scenes now benefit from AI Light as well, where they didn't before.

[mykola1985]

AMAZING! So the new version is coming?

[haze]

AMAZING! So the new version is coming? :-P

Yes, we are aiming XB4 for the next couple of weeks

[Goldorak]

Quick question...

If NETWORK RENDERING is currently disabled, what about the ORC ? If I buy an ORC subscription, does the ORC currently support this v4.00-XB3 build, or will it only work with the official commercial releases like 3.08 ?

I'd stick with 3.08 for ORC. RE is being worked on, both in terms of our own portable version and licensable ones out there too.

[nildoe]

Hello...has offline license been re-implemented yet?

If not any ideas how long it will take? I'd very much like to start using 3.08 or even 4.0

[Hesekiel2517]

Did anyone get AI Light working? I tried it on multiple scenes. The results are always worse (more noise/longer rendertimes) than without it.

Wallace has made some major improvements in this area for XB4, a lot of scenes now benefit from AI Light as well, where they didn't before.

Sounds great! Thank you for the quick answer

[Notiusweb]

I, or my PC rather, has found this release buggy. Buggy as in little weird things freeze the whole PC. Not consistent but:
(1) Denoiser use
(2) Saving image
(3) adjusting a slider

Here is another strange anomaly:
In a list of materials in the scene object menu, you can rename them, If it is a Diffuse, the name change stays.
If a glossy one, the name change disappears. It sometimes will let you type it in but keeps disappearing immediately when you hit enter or click on something else.
Other times clicking where you should to change the name does nothing.

[Notiusweb]

To all the Octane developers - is THIS possible:

BACKGROUND
So when I render an animation, it appears that the denoiser is moving from scratch, frame to frame with each frame, in order to collect the information for the denoise.
So, if I can render a batch of denoised images @1080p for say, 6 seconds per frame, a 900 frame render winds up being 5400 seconds, so like 90 minutes, an hour and a half.

IDEA
What if, the denoiser could run an initial 'Scan Pass' of some sort, where it could scan all frames in a timeline, to gauge AI-wise where the renders wind up being less dynamic and more static (ie 'stay the same')
Then, it could render frame by frame, but not be pressured to start from scratch in all frame cases, because it would have a que of info to cross-check against, (either in a buffer of active info, or off of a text-file, say) in which case it could spit out the almost identically same render, much like if it was a single mat HDRI, or single JPG background. Now, it would need to collect the info first, so it would take time to do it, initially, before render.
But, if it could run such a Scan Pass to collect the info first, it would have a more optimized 'render plan' for the whole project.

HOPE
instead of having to run an image by image denoise scan, it would do one at beginning (ie 15 minutes scan of all frames), que the info, and then run the renders with the AI of knowledge already captured (45 minutes, say).
I am imagining such a Scan pass could cut down the render timer for a sequence of frames.
It doesn't make denoiser or renderer itself faster, it just gives could give 'the project' a curriculum to work off of, AI wise. And then the project could take an hour, vs and hour and a half.

EDIT
--Nutshell--
The pre-scan could identify and tell Octane if any duplicate, or near-duplicate frames, which could then be processed more quickly by Octane, given that Octane will in-advance know it will have already rendered the identical, or near-identical, frame before it.


Any thoughts?