Hi there,
in the docu i read:
—–
"Quantization “
The storage type for the main image. The type of quantization used will affect image quality and size. If you need to adjust the image’s dynamic range in compositing, you should normally leave this value at the default of 16-bit floating point.

The default is ”float16“ for the first plane, and ”float" for secondary planes. You can override the first plane’s value with the -b command line argument to mantra.

“Gamma”

—–
now we can choose:
8bit integer
16bit integer
16bit float
32bit float

when writing png files i get:
8bit integer => an 8 bit png
16bit integer => a 16 bit png
16bit float => an 8 bit png
32bit float => an 8 bit png

but
8bit integer => an 8 bit tif
16bit integer => a 16 bit tif
16bit float => an 32 bit tif
32bit float => an 32 bit tif

I am confused….

In use: Houdini Version 15.5.523

Bug? Feature?

Tif almost behaves as expected appart from 16bit that creates a 32 bit file. And tif is able to save 16bit…

Can somebody please explain this.
I also have the feeling that this behauvior is not persistent…

Apart from that in the help there is nothing about Gamma.

thanks for your help.


Olaf

google drive links to the renders and my test scene:
images:
https://drive.google.com/file/d/0B5Hj67_lrrlkUi12dUVfOTIwdDA/view?usp=sharing [drive.google.com]
scene:
https://drive.google.com/file/d/0B5Hj67_lrrlkRmVzUU5jUk9qLVU/view?usp=sharing [drive.google.com]

A lot of this is driven by what a particular format supports and what can be widely read by common readers of each particular format. To truly get all the formats, you would need to render of Houdini's proprietary image format, .pic. For most things these days though, output to .exr (which is 16 bit float) is the most common (with a gamma of 1.0).

Hi Edward,
thanks for your answer.

Sure exr is a great format and for my normal jobs i use it. I would normaly never output to 8bit directly and convert afterwards.

Now the question stays is the behaviour with png as designed or is it a bug? Do you have the same result?
Could you please try to render to png 16bitfloat? Do you get a 8bit png too?

thanks

Olaf

Olaf Finkbeiner
Now the question stays is the behaviour with png as designed or is it a bug? Do you have the same result?
Could you please try to render to png 16bitfloat? Do you get a 8bit png too?

Hi Olaf,

This is by design.

PNG does not support 16-bit float storage. Only integer values.

Many pieces of software don't properly support 16-bit integer png files. From the spec, it's also not clear what color space 16-bit png files should be in.

Many novice users were being bitten when we quantized the floats to 16-bit integers (even I was bitten by this). If you want a 16-bit png file, you can still specify a 16-bit integer quantization by mantra.

Hi Mark,
cool. Thx.
Has this changed lately? I tested this a while ago …
This makes sense now.

kind regards

Olaf

Olaf Finkbeiner
Hi Mark,
cool. Thx.
Has this changed lately? I tested this a while ago …
This makes sense now.

kind regards

Olaf

https://www.sidefx.com/changelog/?journal=15.5&version=15.5&categories=&build_0=468&body=&build_1=468&show_versions=on&show_compatibility=on&items_per_page= [sidefx.com]

Hi Mark,
next time something confuses me i know where to look!

Thx for the great help and support.

Olaf