I should add, I just checked the speed with each channel split off and read
in separately then copied back to the same multichannel stream inside a
Group node, it's about 10-15 times faster so we will definitely be going
down that route. It's something I've taken for granted working at various
places where you get the auto compiled group node but I didn't realise the
difference was so huge compared to a true multichannel image.



On 8 May 2013 14:52, Michael Garrett <[email protected]> wrote:

> Piotr, thanks for confirmation straight from the source.
>
> Elias, these files have 16 bit rgba but all other channels are actually 32
> bit. But keeping in mind that this is with Mantra.
>
>
>
> On 8 May 2013 14:41, Elias Ericsson Rydberg <
> [email protected]> wrote:
>
>> If I'm not misinformed you should be able to mix 16 and 32 bit in one
>> single exr. However I think your rgb layer have to be 32 for this to work.
>> At least this is my experience with Arnold for XSI.
>>
>> 8 maj 2013 kl. 20:37 skrev Piotr Stanczyk <[email protected]>:
>>
>> Hi,
>>
>>
>> OpenEXR 1.6.1 and 1.7.0 do not differ in that respect. v1.6.1 could store
>> channels of different types in the file. How applications treat them is
>> quite another matter though ...
>>
>> Piotr
>>
>>
>>  ------------------------------
>> *From:* [email protected] [
>> [email protected]] on behalf of Michael
>> Garrett [[email protected]]
>> *Sent:* 08 May 2013 11:28
>> *To:* [email protected]
>> *Subject:* Re: [Nuke-users] Re: EXR file size Mantra vs Nuke re-render
>>
>>  I just had another look in exrheader and indeed, a bunch of these
>> channels are 32-bit, although the beauty is 16 bit and Nuke sees the file
>> as 16 bit in the metadata.
>>
>>  Would this be because Houdini is using OpenEXR 1.7, which is backward
>> compatible? As far as I knew, 1.6.1 was only able to store a single data
>> type across all channels. I'm interested to know, but at least I'm aware of
>> the issue now.
>>
>>  Thanks for bringing this to my attention by the way!
>>
>>
>>
>> On 8 May 2013 11:22, Michael Garrett <[email protected]> wrote:
>>
>>> Interesting...I had thought that the ability to have a 32 bit channel or
>>> layer in an otherwise 16 bit half float file was something you could only
>>> do with OpenEXR 1.7.
>>>
>>>  I know it's not best practice to have multichannel exr's right now,
>>> I'll look into the possibility of getting them broken out as well.
>>>
>>>  I will double check with the lighter as to what her output settings
>>> are, but both the metadata tag and exrheader are telling me it's zip1
>>> scanline ((aka exr compression = 2).
>>>
>>>
>>>
>>>  On 8 May 2013 00:00, marty b <[email protected]> wrote:
>>>
>>>>  **
>>>> Good call John.
>>>>
>>>> Mantra render with 16bit 'C' & 32bit extra planes = 2.5Mb
>>>> Mantra render with 16bit 'C' & 16bit extra planes = 1Mb
>>>>
>>>> Re-render in Nuke at 16bit = 760KB
>>>>
>>>> In Mantra just set the Quantize of the Extra Image Planes to 16bit
>>>> float from 32bit float.
>>>>
>>>>  _______________________________________________
>>>> Nuke-users mailing list
>>>> [email protected], http://forums.thefoundry.co.uk/
>>>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>>>
>>>
>>>
>>    _______________________________________________
>> Nuke-users mailing list
>> [email protected], http://forums.thefoundry.co.uk/
>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>
>>
>> _______________________________________________
>> Nuke-users mailing list
>> [email protected], http://forums.thefoundry.co.uk/
>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>
>
>
_______________________________________________
Nuke-users mailing list
[email protected], http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users

Reply via email to