I was wondering if anyone is working with higher bit color depth and what would the advantage be, and do we actually need to set it high, when we are not doing any printing?

I usually just use the standard 16 bit, with the 32 thrown in when I have an alpha, but would there be any reason to go higher, when doing things like textures?

the only reason i've found to go for anything higher than 8bit per channel was when using displacement or converting height to normals to minimize banding. authoring "normal" textures in more than 8bit per channel seems like quite a bit of ressource waste imho.

Yeah, same here, and it slows down photoshop as well. I was thinking it might work if you want to do a large gradient, but would think that you still would get banding when converting back to 8 bit.

yeah, you definitely will, but the banding pre-conversion won't get treated as an actual edge that should be a proper normal edge.

According to crytek, you should ALWAYS be authoring in 16 bit.

technical deets here.

I vouch for 16 bit too, but sometimes, some engines will spit errors at you if you do try that, so always try and work in a 16 bit environment, but be prepared to revert back to a 8 bit in your final steps.

Btw, I noticed in photoshop that some filters actually seem to not work in higher colors, they are grey out, any idea why that is?

The one cool thing about higher range images is how nicely they behave when it comes to making highlight adjustments. There is something quite magical about adjusting the contrast of a render and seing the brightest and darkest areas behaving so well. Very cool stuff

I would think this stuff also comes in handy when putting together complex image composites and matte paintings too.

Originally Posted by McGreed

Btw, I noticed in photoshop that some filters actually seem to not work in higher colors, they are grey out, any idea why that is?

http://forums.adobe.com/message/2943383

that pretty much answers that

the idea with 16bit for normalmaps is to avoid having to compress them twice; once to 8 bit (per channel! 24 bits per pixel) and another to dxt5 (1 bit per pixel). obviously compression will only get worse if it's compounded.

i usually work in 16bit when dealing with procedural gradients. you can do so much more with so much less banding issues, especially in the darks.