LuxCoreRender Forums

lacilaci wrote: ↑Mon Oct 08, 2018 3:12 pm

lacilaci wrote: ↑Mon Oct 08, 2018 2:48 pm

marcatore wrote: ↑Mon Oct 08, 2018 2:40 pm Krita uses 32bit per channel. As suggested should solve the problem

It wouldn't let me import 16 bit tiff (from blender)... gonna try exr

Ok, exr works, not sure why tif doesn't...

I think that TIFF 16bit is not really an HDR format: the values are just an integer between 0 and 2^16-1 instead of [0, 2^8-1] while .exr uses 16 bit half (i.e. 16bit half floating point numbers: https://en.wikipedia.org/wiki/Half-prec ... int_format).

Dade wrote: ↑Mon Oct 08, 2018 4:03 pm

lacilaci wrote: ↑Mon Oct 08, 2018 3:12 pm

lacilaci wrote: ↑Mon Oct 08, 2018 2:48 pm

It wouldn't let me import 16 bit tiff (from blender)... gonna try exr

Ok, exr works, not sure why tif doesn't...

I think that TIFF 16bit is not really an HDR format: the values are just an integer between 0 and 2^16-1 instead of [0, 2^8-1] while .exr uses 16 bit half (i.e. 16bit half floating point numbers: https://en.wikipedia.org/wiki/Half-prec ... int_format).

I know, but I don't really need 32bit in my workflow. I just used some basic chromatic abberation and subtle contrast adjustment. 16bit is pretty much enough for that. Doesn't matter im gonna use exr from now on

lacilaci wrote: ↑Mon Oct 08, 2018 4:14 pm I know, but I don't really need 32bit in my workflow. I just used some basic chromatic abberation and subtle contrast adjustment. 16bit is pretty much enough for that.

The numbers expressed by a 16bit integer and by a 16bit floating point are very different: one can not represent a value like "0.5" while the other can.

lacilaci wrote: ↑Mon Oct 08, 2018 3:29 pm quicktest (~10min)

start to look great keep pushing please

Dade wrote: ↑Mon Oct 08, 2018 4:48 pm

lacilaci wrote: ↑Mon Oct 08, 2018 4:14 pm I know, but I don't really need 32bit in my workflow. I just used some basic chromatic abberation and subtle contrast adjustment. 16bit is pretty much enough for that.

The numbers expressed by a 16bit integer and by a 16bit floating point are very different: one can not represent a value like "0.5" while the other can.

Well, the way recent blender versions work including 2.79b Tonemapping is handled within blender using aces filmic. so luxcore in-camera settings combined with filmic blender settings get the job done at 90% without user input. Rest is just post effects like chromatic abberation and contrast/curves. Output at 8bit is good enough(for internet/web viewing) I only have the issue cause I want to use nvidia denoiser, that means currently outputing .exr at 32bit float, denoising in standalone using hdr option, then re-importi g to blender for filmic tonemapping/compression and again output or process in blender for some contrast/curve adjustments.

I know I could do CA with luxcore camera settings and its a great feature to have in-camera but I cannot use that as input for denoiser as the more are pixels affected by CA the more are they ignorer or they confuse the denoiser.

Sharlybg wrote: ↑Mon Oct 08, 2018 4:58 pm

lacilaci wrote: ↑Mon Oct 08, 2018 3:29 pm quicktest (~10min)

start to look great keep pushing please

Thanks , there is a lot to do still though.

small update

note to self: do not change min. epsilon value when working with volumes

some more

a pink pool float appears...

aaand from this angle pillows look weird and juice is super dark again... wtf