[FFmpeg-user] maintaining full swing when encoding in 10 bit (want to preserve all values from 0-1023)

Marwan Daar marwan.daar at gmail.com
Wed Nov 11 07:16:58 CET 2015

I tried the scale command with the 10 bit build, but no luck.  Resulting 
video was one uniform color rather than 8 distinct bars.

Next, I tried Zeranoe's new 8 bit build (20151109-git-480bad7) which 
incorporates zimg, and had the following results:

Code used:

|ffmpeg -i 1.png -vf zscale=r='full' 1.mp4

|When the png file is 8 bit (any 8 bit png), I get the following error:

|[Parsed_zscale_0 @ 00788720] code 3074: no path between colorspaces|

However, when the png file is 10 bit (any 10 bit png), no error occurs, 
although the resulting video of course is 8 bit.


> what happens is that there is an auto inserted scale filter used to
> convert from RGB to YCbCr, the auto inserted defaults/features don't
> work for your case you could try explicitly add the scale using
> -vf scale=in_range=XXX:out_range=YYY:out_color_matrix=bt709
> Replace XXX and YYY with 'full'/'tv' and you can control the scaling
> (obviously you should fill in the appropriate matrix values for your
> case).
> It isn't perfect, and depending on your destination you may not get
> full [0-1023] due to some codes being reserved for sync in video
> formats, but this way you may get something closer to what you need
> without needing the recompile.
> Kevin
> On Mon, Nov 9, 2015 at 7:08 PM, Marwan Daar <marwan.daar at gmail.com> wrote:
>> thanks Paul. I'm not sure how to do that though. I found this:
>> https://github.com/sekrit-twc/zimg
>> but have no idea how to incorporate this into my workflow. Do I somehow need
>> to compile this library into a new build? I've never done such a thing - is
>> it easy to learn?
>> -Marwan
>>> That are swscale bug(s). Compile with z.img and use zscale.

More information about the ffmpeg-user mailing list