New rendering approach with the EVO (store rendered result in samples)

[jasonhuang1115]

Hi,

I came across this interesting new rendering approach to store result in samples instead of pixels. http://www.go-ghost.com/
From descriptions, seems really powerful with smaller file size than Deep images.

Found Chaos Group is mentioned on their "About Us" page. Can we learn more about if this is the direction V-Ray is heading or all things are still under NDA?

cheers,
Jason

[vlado]

We have an NDA with these guys, so I can't share much, but you can contact them for more information.

In my personal view, their approach is what sub-pixel compositing should have been from the start. Deep images and OpenEXR 2 missed their mark slightly in that regard.

Best regards,
Vlado

[Vizioen]

Looks amazing. Those are soms nice prospects.

[instinct]

Mail sent. While i really dig the approach some things seem odd (e.g. the claim that deep can not accurately seperate objects. It can't with z-based sample merging, but vray also does renderID based merging and with that it can). Would really love to hear more. Will see if they get back to me.

Thanks for the heads up,
Thorsten

P.S. it's not a new rendering approach, but a new file format and storage approach

[^Lele^]

Oh, i didn't know about this.
It vindicates me not WANTING to get down to the nitty-gritty of deeps (I always thought of it as inferior to the old RPF, so i didn't touch it with a stick.).
Can't wait to test this stuff out.

[vlado]

It can't with z-based sample merging, but vray also does renderID based merging and with that it can.

Yes, we can do that, but if two points have the same z-depth but belong to different objects, there is no good way to represent that into a regular deep file. We do a trick where we detect these cases and nudge the z-depth a little bit, but it's an ugly hack.

Best regards,
Vlado

[instinct]

Yes, we can do that, but if two points have the same z-depth but belong to different objects, there is no good way to represent that into a regular deep file. We do a trick where we detect these cases and nudge the z-depth a little bit, but it's an ugly hack.

Which also comes into play for other cases. Got it. I am really eager to learn if/how they are also handling secondary samples.

Cheers,
Thorsten

[Art48]

Now that does indeed look really interesting. Looking forward to it!

[vlado]

Specifically the up-scaling functionality is very interesting...

Best regards,
Vlado

[Art48]

What doesn't get into my head is why the file size would be smaller - wouldn't they have to store thousands of times more with all the samples or did I misunderstand it?

[vlado]

What doesn't get into my head is why the file size would be smaller - wouldn't they have to store thousands of times more with all the samples or did I misunderstand it?

That I don't know. We'll have to play with it and see how that works.

Best regards,
Vlado

[Syclone1]

Just looked at their website and after cleaning up from drooling I sent an email for more info as well

[jasonhuang1115]

Thanks for the reply Vlado. It's exciting to see what it can do with "promised" smaller file size. Will see how the industry react to this new format.