I am wondering, has anyone had a lot of experience with using texture maps when rendering with HDRi? If so, does this produce good results even though the texture will already have it's own lighting.

i.e. Taking a picture of bricks (fromt the real world) and using that as your texture map on a piece of geometry and then HDRi lighting that geometry.

Just curious….maybe there's a trick to it…

Cheers!

Anytime you use texture maps that have their own inherent lighting you'll be fighting that no matter what sort of lighting approach you take. If it's distinctive enough - it will never look good. Better approach is to try to get neutrally lit tmaps or better yet, procedural shaders.

Cheers,

J.C.

not really on topic but rather than create a new one…

do i have to convert hdri maps to rat format as the video tutorial suggests? they seem to work as hdr but maybe they dont work properly

i assume i have to use vertical cross format? these dont seem to be very common (can convert with hdrshop i believe.. so not a big issue if so..) but i actually tried to isixpack some vertical cross ones and i got an error.

does houdini support openEXR fully now? by that i mean 24 bit.. suitable for hdri substitute… is there an advantage of openEXR over hdri? for everyday use i mean…

do the EXR have to be probes - i.e. made for lighting? i downloaded the sample images from lucasarts, but they seem to be standard images, not ready for use as a light probe…

just some questions ive thought of recently feel free to answer some, all - or none, cheers

do i have to convert hdri maps to rat format as the video tutorial suggests?

Don't have to - hdr can be read and used in mantra directly, however there are indeed advantages to converting to rat - for one thing you can filter the map in the texture call with rats - not so with other formats. Personally I tend to use rats everywhere, although now that exr is here…hmmmm…

but i actually tried to isixpack some vertical cross ones and i got an error

from the isixpack help:

Alternatively, it's possible to take a single image which is laid out in
the shape of a cross (similar to http://www.debevec.org/Probes/) [debevec.org] and
generate a single reflection map (must be a cubic .rat file).

does houdini support openEXR fully now? by that i mean 24 bit.. suitable for hdri substitute… is there an advantage of openEXR over hdri? for everyday use i mean…

Houdini, as of the last few cuts, does support exr completely throughout the pipeline - compositing, mantra, etc. There are four very kewl things about exr:

1. ILM determined through experimentation and real world production that using this “half” format(16 float instead of 32) gives you results that are pretty durned hard to differentiate from pure 32 bit floating point. It's actually significantly *better* than 24 bit - it's floating point. I've been fiddling with it a fair bit the last few days and it's amazing what you can get rendering oexr layers for some fancy comp footwork. I've actually had to watch myself, however, since in the past I might do some sloppy tricks that counted on, say, an alpha channel peaking at “1” and not going any higher. Now there's headroom , and I had problems down the pipe with the comp. I actually have to be more careful now - assume nothing! However, the freedom you get, the comps you get - just being able to pull brightness by 1/2 and not get that gruesome flat grey…yummy!

2. 16 float = half the size of 32! Quicker to render! Zoom!

3. Lots of other nicieties in the oexr spec, like not just 16 - but the full range of bit depths, mip mapping, scanline, dataWindow/displayWindow(which lets you have an image that is larger than the actually area you are going to see - really handy for compositing down the line), etc. Of course, Houdini doesn't yet support *all* these features(notably the last one), but hopefully…

4. In cahoots with NVidia, the 16 “half” format has some very clever display tricks with high end NVidia cards. I can display an oxer image on my FX3000, and it moves around almost as fast as an 8 bit! Try that wth 32 bit and it will crawl…

So definitely, yes, oexr support is really there - highly useful, and it's fast.

thanks JC, this makes some sense. So EXR is not just a different format of floating point image! it has other features, like rats:

JColdrick

you can filter the map in the texture call with rats

i read somewhere that amongst other features, you dont need to worry about antialiasing with your textures when using rats (and i assume EXR)? is this what you meant?

anyway i think i get the point - they are even more than just high dynamic range images.

this maybe RTFM - i need to do more research is there an hdrshop equivalent (linux if possible) for EXR? the ILM site seemed linux friendly with .tar.gz file extensions on their downloads. anyway ill read up some more.

cheers!

No - not hdrshop. Really, Oexr is simply another file format, like tiff, targa, etc. It just happens to be a new one that can encapsulate a wide range of options typically used in the film post FX world, plus it's pushing the idea of the “half” float. Plus ILM has been using it for production, so it's pretty robust. There's a bunch of other clever things in there too, you can read about it on the OExr site…

Getting “deep” images from the real world using 8 bit capture still involves fiddling about with hdrshop-style methods. However, you can render your CG elements in Houdini right out to OExr, with all the benefits. Try rendering a sphere with some procedural noise on it, preferably a very bold, contrasty regular 24 bit integer tiff. Don't use texture maps(since the “deepness” of the result will be driven by the 8 bit image!). Now, change the call to mantra in the ROP to “mantra -b half”, and change the file out to an OExr file. Pull both of these into the compositor. They'll look identical by default. You should notice when you MMB on them that one is 8 bit, the other 16 Float. Now, append a Bright COP to each, and start pulling down the brightness to around 0.5. Hopefully, you should notice that the hotspots of the 8 bit become a dull grey flat colour, while the 16 bit Float should retain a hotspot(although overall it will become darker, it still has “punch”).

Somewhere to start fiddling, anyway…

Btw, if you're using the Apprentice version - i don't think all this is there yet. new one coming out soon - try it there.

Cheers,

J.C.

yep apprentice user, so ill wait for the next release.

i suspect we are thinking about this on different levels - to me it is no different to standard HDRI - but i have begun to seen some advantages to EXR (lossless compression!)

i imagine a way to make EXR light probes will eventuate for image based lighting, ill stick with hdri for now, no probs

sample 4 on this page is impressive: http://www.openexr.com/samples.html [openexr.com]

thanks

We're not looking at it differently. A “light probe” is just a high dynamic image, with a particular preset group of mappings(spherical, etc.). The whole business of how you get information from cameras that are designed to capture 8 bit integer information into something that is much richer, is the whole hdrshop thing. It's just a methodology. OExr is no different - the biggest advantage is that you can get some pretty great images at a *much* smaller source file size. Also, it's designed to work really well on the compositing side, not just as texture maps.

But yes , for now, getting real world imagery into these “deeper” maps is easiest with hdrshop - then convert later…

Cheers,

J.C.