Blender and Ultra High Definition Television (UHD)

Support Technical Support
1

Hi,
I didn’t see a mention of UHD on the 2.7 - 2.8 roadmap of Blender (http://code.blender.org/index.php/2013/06/blender-roadmap-2-7-2-8-and-beyond/)
or on plans for 2014-2015 (http://code.blender.org/index.php/2013/12/plans-for-2014-2015/).

With Blender you can already output pictures 4K size (3840 pixels wide by 2160 pixels) or 8K size (7680 pixels wide by 4320 pixels).
But as far as I know Blender is limited to 8 bits per color channel (the UHD norm is 12 bits), and there is no High Efficiency Video Coding (HEVC) output (nor 22.2 surround sound).

Is there a plan to integrate UHD requests in Blender ?

2

Blender is not limited to 8bits per channel, it is limited to 32bit float per channel… more than enough room for UHD…

3

In the color wheel you can enter value (in hexa) from 000000 to FFFFFF, that seems 8 bits per color for me.
How can you enter a value from 0 to 4095 per color ?
Are pictures with 12 bit per color well managed in compositing / VSE?

4

you can enter values higher than 1 on the rgb sliders. hex, as far as I know, doesn’t support more than 8-bit values

5

Thanks for the tip, but it’s not OK: I made a quick test (a plane with a color ramp and emission shader) the value of the result are out of range (more than 1).
See the doc http://en.wikipedia.org/wiki/Rec._2020 for the value: for 12 bits, the range between 256 and 3760 are visible colors (from 0.0 to 1.0).

6

Yes, you should be able to convert to Rec.2020, but the internal working of blender is far higher then the bit depth needed for this format.

32bit float… it means that not only is there 32 bits to describe the colour… but the decimal place is floating… 32bit int gives about 4million colours per channel (http://www.spherevfx.com/written-training/miscellaneous-written-training/file-formats-channels-and-colour-depth/ ).

The RGB sliders… dont enter hex, that will be 8 bit… the rgb values yes can go higher then 1, 1 is white… but it can also have a decimal point… that decimal point can be quite accurate.

just because the range for rec2020 is between 256 and 3760… just means that 256 is mapped to the value 0 in blender and 3760 is white.

7

Moved from “General Forums > Blender and CG Discussions” to “Support > Technical Support”

8

Blender is a scene referred 32 bit float model internally.

By default, there is an overly simplistic display referred transformation 1D LUT on the output, but this in no way is absolute.

If one stacks several point lights in front of a cube, the cube will look quite blown out to white. In a scene referred model however, there is no such notion of white. One could choose to do two things to achieve this transform to the display referred domain:

  • Adjust the white and black point scaling in the Color Management control in the Scene panel.
  • Use a different display referred transform on the output.

UHD is a display referred format, and requires display referred values in a unique colorspace.

With a little care and knowledge, one could easily deliver both from within Blender, suitable for encoding.

With respect,
TJS

9

What this 8-bit number represents is the base colour, the colour under ideal lighting conditions. If you were entering each step of colour change as you went further and further around the ideally-lighted object into shadow or into highlight, then there would be cause for concern.

And I don’t think there’s much advantage in getting any more picky about the base colour than Blender already is. On most monitors, if the monitor isn’t calibrated properly for the room’s lighting environment, you’ll be hard pressed to get that accurate anyway.