Have anyone ever managed to do cryptomatte AOV reader in SD?
Is it at all possible ? If I understand it right Cryptomatter writes/stores somehow multiple ID color values and its weight into same pixel.
I never managed to learn how further extraction goes exactly. My guess it should be using some range separation or something to extract masks ( i have no idea really)
Does anyone managed to do it in SD? is it possible it all in its pixel processor? I once tried a typical idea of combining depth and color value into same 32 bit bitmap pixel and then splitting depth out and it haven't gone very well showing me some weird artifacts ?
Replies
there's no reason you can't combine depth and color into the same image - you've got 4 channels, they're not compressed provided you're using a sensible format (i.e. exr) so you shouldn't see any channel crosstalk or corruption of the data
if multiple channels are packed into a single channel it will be a problem - designer doesn't have tools for bit manipulation so encoding/decoding this sort of thing is kind of not possible (i think it is possible, I just haven't managed to do it)
if you want to post some sample data up here and explain what data is stored where i can have a poke at it - its something I've wanted an excuse to look into for a while
It may be possible to generate a substance graph to do it but I suspect youd have to make a new one for every different image.
If you need masks out of it I'd suggest generating them in your renderer
It's been a year, anyone got any new news on Cryptomatte for substance?
I very tired of having to re-setup my ID map selections when there is a model update and I need to re-bake. Cryptomatte stores object data and will not forgot what mask layer its assigned to unless the object names change. This is immensely helpful in working in tandem with modeling during production.
I'm not a programmer so I'm kinda of relying on the community for this one :D