[FFmpeg-user] Support for Prores4444 12bit files

Robert Krüger krueger at lesspain.de
Thu Jul 4 19:10:18 CEST 2013


On Jul 4, 2013 6:30 PM, "Paul B Mahol" <onemda at gmail.com> wrote:
>
> On 7/4/13, Robert Krueger <krueger at lesspain.de> wrote:
> > On Thu, Jul 4, 2013 at 5:25 PM, Paul B Mahol <onemda at gmail.com> wrote:
> >> On 7/4/13, Robert Krueger <krueger at lesspain.de> wrote:
> >>> Hi,
> >>>
> >>> Apple's whitepaper states that Prores4444 has "up to 12bit" color
> >>> depth. There are no Apple tools that I know of that have a switch to
> >>> produce such a 12bit beast (their professional transcoder "Compressor"
> >>> has the options "16.7 million colors" or "more than 16.7 million
> >>> colors" which is useless in that case as IMO that translates to 8bit
> >>> and more than 8 bit).
> >>>
> >>> Now there exist recorders which explicitly claim to support 10bit
> >>> _and_ 12bit Prores4444 (see
> >>> http://www.sounddevices.com/products/pix240i/sample-files/) and I
> >>> downloaded those and tested transcoding with ffmpeg and the output
> >>> looks fine. Since the pixel formats used by ffmpeg for Prores are 10
> >>> bit only, I am wondering if there is a way to test if the color depth
> >>> is really 12bit and thus reduced in any transcoding.
> >>>
> >>> Could anyone with more intimate knowledge of the decoder(s) shed some
> >>> light on this? Is the current decoder reducing bit depth of these
> >>> files when decoding and if so are there plans to add support for a
> >>> pixel format that allows for 12 bit?
> >>>
> >>> Is there a way of testing such a file if it is 10 or 12bit? Apart from
> >>> the alpha channel (which in this case is not there) neither the header
> >>> documentation at
> >>> http://wiki.multimedia.cx/index.php?title=Apple_ProRes nor the source
> >>> code reveal anything that would be suitable for making that
> >>> distinction? Or is the formulation "up to 12bit" simply wrong and
> >>> Prores4444 is always 12bit?
> >>>
> >>> Best regards,
> >>>
> >>> Robert
> >>>
> >>> P.S.: I am not talking about 10 bit files with a special gamma
> >>> function like Arri Log-C or Sony S-Log but real 12bits of depth.
> >>
> >> I think you are confused. There is no 12bit prores.
> >
> > OK, so your theory is:
> > - Apple's official whitepaper is bullshit
> > - http://www.sounddevices.com/products/pix240i/specs/ + a lot of
> > articles about it are fraud or just bullshit
> >
> > Definitely possible though it does not sound too likely and i am
> > trying to find out. But you are correct. I am indeed confused.
>
> Looking at both obfuscated decoders there is possibility
> prores can store any bit depth.
>
> If they decode real 12bit as 10bit and output is same as given by
> other decoders that claim they support real 12bit than I'm
> in wrong universe.
>
I am looking for a way to really compare the decoder output because
visually it is not possible to distinguish 10 bit from 12bit (unless you
manipulate it in a color grading application that is known to use the full
12bit). I probably have to decode them using libavcodec and using apple
apis (where you can decode it to a16bit pixel format) and then compare the
numbers if I don't find an easier solution. Furthermore I am trying to get
a sample from a more official (preferably Apple) source. There is always
the possibility that the sample on the site is not really 12bit because
someone messed up.


More information about the ffmpeg-user mailing list