AO with the render itself as a Pass

[thomaskc]

So when using the extraTex method do you fill in the background manually to apply the AO in post? because to me it seems like fixing up the map afterwards for it to be useful, takes just as long as submitting a quick override job with your initial render?

[NicoC]

Well, the extraTex method is fine for me - all the more that we use PSDManager for output, so the ExtraTex RE layer is already set up.
I'm just somehow nostalgic of a finalRender feature allowing to output AO (in GI tab) and/or Detail Enhancement directly in the AO RE ... you just need to check Output GI details (or whatever the name is) option in AO RE to get them separately. Pretty straight forward, isn't it ? :P

[m_hinks]

I remember asking/seeing a request for the extraTex RE to have a user selected bg colour, ie White so fringing is not an issue.

[prateekvishwa]

I remember asking/seeing a request for the extraTex RE to have a user selected bg colour, ie White so fringing is not an issue.

tht will be awesome!

[joconnell]

I remember asking/seeing a request for the extraTex RE to have a user selected bg colour, ie White so fringing is not an issue.

Why not just swap your occluded and unoccluded colours so you've black on black and then no fringe? It'd mean an invert in post but that's not the worst thing in the world?

[Rens]

What joconnell said should work, also when you're using a proper un-premultiply like nuke has there should be no issue with fringing, well, at least on the alpha edges. : )

Also make sure you've disabled 'consider for antialiasing' unless you really need it! It's a huge difference as it uses the AO contrast as well to determine where to use more samples.

Using an AO-extraTex pass like that is relatively error-free and wouldn't take a huge hit on render times.

[m_hinks]

Why not just swap your occluded and unoccluded colours so you've black on black and then no fringe? It'd mean an invert in post but that's not the worst thing in the world?

Indeed a good solution, thanks John.

Something to note, and I have asked Vlado/Choas if it can be changed, is that glass will render as a solid object in the AO, ie objects the other side of the glass will not render with AO.

Also a bug is that if you set the glass materials refraction to affect all channels (which you would have to do to get multimatte masks through glass), the white of the AO over glass will render very high float values, where other flat objects that show as white are float 1. I touch annoying, but you can apply the AO pass in nuke etc as an 8 bit image which fixes the issue.

[Rens]

Also a bug is that if you set the glass materials refraction to affect all channels (which you would have to do to get multimatte masks through glass), the white of the AO over glass will render very high float values, where other flat objects that show as white are float 1. I touch annoying, but you can apply the AO pass in nuke etc as an 8 bit image which fixes the issue.

Isn't it just adding the the AO values as it would add the multimatte values?

[tehamnes]

A bit on the side of this interesting topic:
Noone using the AO option in GI?
I mean if you know you want the AO. Just curious if there are other drawbacks besides that you cannot fiddle with it in comp or have the image without AO. Seems like Extratex with Vraydirt adds more rendering time and is less accurate? + the time needed to comp. it in afterwards...