Rendering openEXR "half"

Become a member of the CGSociety

Connect, Share, and Learn with our Large Growing CG Art Community. It's Free!

Thread Tools Display Modes
  09 September 2013
Rendering openEXR "half"

When I render to openEXRs, I've historically always used one of the first 5 compression methods and stayed away from the lossy 16-bit options.

I've always thought 32-bit float = awesome. Anything "Lossy" = bad.

But that doesn't seem to be the case with floating point color. As I'm using Nuke more and more, it seems like almost everything is 16-bit "half" float (it's the default format on the 'write' node). Nuke's documentation recommends 16-bit float as it has less overhead that full 32-bit float. And After Effects renders out 16-bit float EXRs by default. (The AE ProEXR plug-in maker recommends against writing full 32-bit float files because it is overkill)

From my readings, 16-bit float (aka "half") contains more than enough color data for high-end production/film/vfx work.

Love to hear other people's opinions and experiences with 16-bit float.
----end of line
  09 September 2013
We've been using a VRay > half float EXR > Nuke workflow for the last two years or so. It works very nicely, to the extent that I use Nuke over Photoshop when I can now. File size and disk read / write speed have been more important than any possible loss of quality (not that I've ever noticed any).

I believe there are some instances when full float is important, for example with data passes (position pass, motion pass). I've not had a problem with using half float for these but I'm not doing crazy hardcore comp work.
  09 September 2013
Youve pretty much got it, lossy formats are perfectly fine so long as they retain enough data for you to not notice.
Matthew O'Neill
  09 September 2013

As far as I understand it, 16 bit float is more than enough in most cases.

I found this link to a discussion. It had some very good information about the issue.

  09 September 2013
We'll eventually go to a full 32 bit pipeline but fact is right now with current specs, disk space bandwidth and processing are all still better suited for 16 bit, especially as we increase resolution again and possibly frames rates (and stereo)

The way half floats work can be pretty flexible where there's a balance between color and luminance precision. Most use exr and there core the 5 bits of color and 10 bit mantissa for color offering a decent 30 stop range possibly a little higher but with discernible loss in acurracy.

R&H with its historically more coomp centric lighting workflow wrote their own Half float file format that had a larger mantissa offering for even more exposure level accuracy but at the sacrifice of color precision.

The thing is that these are still far beyond the information in a raw photo or film so really we are at a point where the cg image has more than the film it's being applied to so the sacrifice of 32 bit is a minor concern
Quote: "Until you do what you believe in, how do you know whether you believe in it or not?" -Leo Tolstoy
Kai Pedersen
  09 September 2013
Thread automatically closed

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.
CGTalk Policy/Legalities
Note that as CGTalk Members, you agree to the terms and conditions of using this website.
Thread Closed share thread

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Society of Digital Artists

Powered by vBulletin
Copyright 2000 - 2006,
Jelsoft Enterprises Ltd.
Minimize Ads
Forum Jump

All times are GMT. The time now is 12:04 PM.

Powered by vBulletin
Copyright ©2000 - 2018, Jelsoft Enterprises Ltd.