Interesting although a strange conclusion for me.
When you say Alexa ISO 800 is not acceptable for VFX; surely anything is
'acceptable' - even underexposed 8mm... :)
Personally, I'd rather have the stops to play with.
N
Neil Rögnvaldr Scholes
www.neilscholes.com
On 30/04/14 14:43, Adrian Baltowski wrote:
Probably you have access to more recent Alexa but with my footages (2
different, quite recent cameras), with standard ISO 800 and V3LogC
linearisation highlight area (like big fire, chimeras, butterflies
etc.) are clipped on value around 37. There are higher values- around
50- but that are not real "scene pixel" values but just artifacts of
debayer algorithm altogether with sharpening. When I change algorithm
and set sharpen to 0, those exces pixel values just disappears.
It's definitely not fair to say, that Epic is so bad. For me it's just
compromise between noise and dynamic range. With standard settings
Epic has less noise but with cost of dynamic range in highlights. In
my opinion most DPs accept higher noise level from Alexa because it
looks more like film grain than ordinary "electronic" noise from Red.
But for me noise level with the standard Alexa ISO 800 is not
acceptable for VFX, even if dynamic range is better than in Epic. When
you takes it into account in your comparison, difference between them
goes down to around 1 stop instead of 3 I think.
Best
Adrian
W dniu 2014-04-30 01:52:24 użytkownik Schneider, Abraham
<[email protected]> napisał:
Hm, don't really get your point. Of course I linearized both
footages. If I'd compare linear with log, the values are
different. But if I linearize both with the correct curve and have
one footage clipping at values below 5 and the other at 54, that's
a huge quality difference to me.
Of course it would be better to test the cameras side by side at
the same location, but I don't have a RED cam available. That's
why I'm asking here if anyone knows if it is normal with RED to
have clipping at such 'low' values (compared to what I'm used from
Alexa).
Abraham
Am 30.04.2014 um 01:12 schrieb Adrian Baltowski:
Just remember that with all digital cameras these are
arbitrary logarithmic-like curves applied on top of the linear
data captured by the sensor. Currently I'm working on the
Alexa footage: night shot with house on fire. With AlexaV3logC
curve maximum linearized values of fire are around 36 and are
clipped. Just change curve to standard Cineon and THE SAME
clipped highlights goes down to 9, with just little change of
values below 1. And redlogfilm curve is intended to mimic cineon.
So to be fair You need to make tests side by side, with the
same scene captured by both cameras and then compare the results.
Best
W dniu 2014-04-29 20:38:40 użytkownik Randy Little
<[email protected] <mailto:[email protected]>> napisał:
Isnt there a pdlog 685 setting still in red? But it would
have change with redlog and color space if it was going to
change at all I suspect. (Maybe)
On Apr 29, 2014 1:07 PM, "Schneider, Abraham"
<[email protected] <mailto:[email protected]>> wrote:
You're suggesting 'RedlogFilm' in the RED settings, right?
Doing this and using a separate colorspace to do the
Log2Lin gives similar values than using the Read
colorspace. My maximum value in the image is lower
than 5 compared to the > 54 with an Alexa image.
Am 29.04.2014 um 18:17 schrieb Randy Little:
> What happens if you do pdlog in red settings and
then use a colorspace node to do loglin. Read node
loglin does weird things. Colorspace node seems to
work better.
>
> On Apr 29, 2014 11:43 AM, "Neil Rögnvaldr Scholes"
<[email protected] <mailto:[email protected]>> wrote:
> Ha - this is interesting i don't know the answer
but i do know a good DP who hates RED for exactly it
inability to retain highlights.
>
>
> Neil Rögnvaldr Scholes
>
> www.neilscholes.com <http://www.neilscholes.com/>
>
> On 29/04/14 16:33, Schneider, Abraham wrote:
> Thanks!
>
> I assumed all this. Just wanted to make sure there
is no mistake in doing one specific way.
>
> My main question is: is it true that RED cameras can
only deliver/capture highlights that are way darker
then what the Alexa does? So in my example, a maximum
value of around 4 from the RED would be more than 3
stops less than the Alexa with it's value of 54
(linear floating values measured in Nuke).
>
> So the provocative question would be: is RED really
that much worse than the Alexa?
>
> Abraham
>
>
> Am 29.04.2014 um 17:10 schrieb Simon Björk:
>
> Some info on Red footage an it's curves:
>
> - Half Float Linear is the correct linear curve. The
one that says Linear is scaled/offset to be between 0-1.
> - RedLogFilm is a Cineon curve. By applying a
standard log2lin operation, the result is identical to
Half Float Linear.
> - Both Half Float Linear and a linearized RedLogFilm
will produce values below zero. You can avoid this by
converting to PLogLin instead, or just bring up the
values with a standard Add node.
> - Rendering RedlogFilm or linear EXRs from RedCineX
will produce identical result as importing the R3D
directly in Nuke.
> - By rendering out an ACES file from RedCineX you
will get a larger color gammut, but you will not get
any more information in the black/white point.
>
> To sum up:
>
> If the values are clipped when using Half Float
Linear in Nuke, you will not get any more information
by applying different settings or going to a different
application.
>
> Best regards,
> Simon
>
>
> 2014-04-29 16:20 GMT+02:00 Schneider, Abraham
<[email protected] <mailto:[email protected]>>:
> Hi there!
>
> As we're more used to Alexa footage here, I just
wanted to ask about RED, as I have some footage here
that let's me wonder about the capabilities of the RED
camera:
>
> I have some shots with fire, sun, lamps, etc. in my
images. Shot with EPIC-X. When importing them into
Nuke and linearizing them, I always see that these
bright highlights are clipping at around a value of 3
to 4 (depending on the channel).
>
> This happens, no matter how I treat the footage:
> - importing the original R3D file in Nuke and use
"half float linear" as gamma space and 'linear' as the
Read colorspace
> - importing the original R3D file in Nuke and use
"RedlogFilm" as gamma space and 'cineon' (or PlogLin)
as the Read colorspace
> - converting the R3D in RedCine X Pro to DPX with
RedlogFilm and importing with 'cineon' or 'PlogLin'
> - converting the R3D in RedCine X Pro to EXR with
ACES and importing with 'linear'
>
> Having similar takes shot with the Alexa as
AlexaLogC, bright highlights will clip at a value of 54!
>
> I don't want to start a 'war' Alexa vs. RED, but
just wanted to know if this really is 'normal', so the
RED will not deliver higher values, or if there is
something wrong/strange is going on here.
>
> Thanks for your help,
>
> Abraham
>
>
> Abraham Schneider
> Head of VFX pipeline / VFX Supervisor
>
>
> Türkenstr. 89, 80799 München /
> Phone +49 89 3809-1096 <tel:%2B49%2089%203809-1096>
>
> EMail [email protected] <mailto:[email protected]>
>
> Visit us on Facebook!________________________________
>
>
> ARRI Film & TV Services GmbH
> Sitz: München Registergericht: Amtsgericht München
> Handelsregisternummer: HRB 69396
> Geschäftsführer: Helge Jürgens, Josef Reidinger
> _______________________________________________
> Nuke-users mailing list
> [email protected]
<mailto:[email protected]>,
http://forums.thefoundry.co.uk/
>
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
<http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users>
>
> _______________________________________________
> Nuke-users mailing list
> [email protected]
<mailto:[email protected]>,
http://forums.thefoundry.co.uk/
>
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
<http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users>
> _______________________________________________
> Nuke-users mailing list
> [email protected]
<mailto:[email protected]>,
http://forums.thefoundry.co.uk/
>
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
<http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users>
>
>
>
> _______________________________________________
> Nuke-users mailing list
> [email protected]
<mailto:[email protected]>,
http://forums.thefoundry.co.uk/
>
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
<http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users>
> _______________________________________________
> Nuke-users mailing list
> [email protected]
<mailto:[email protected]>,
http://forums.thefoundry.co.uk/
>
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
<http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users>
_______________________________________________
Nuke-users mailing list
[email protected]
<mailto:[email protected]>,
http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
_______________________________________________
Nuke-users mailing list
[email protected]
<mailto:[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
