[FFmpeg-user] Creating ffv1 full-range and broadcast-range files

Andy Furniss adf.lists at gmail.com
Thu May 15 17:47:18 CEST 2014


Carl Eugen Hoyos wrote:
> Andy Furniss <adf.lists <at> gmail.com> writes:
>
>> "I was using in_range and out_range for the test .."
>>
>> Of course yuv420p and yuvj420p without using *range will produce a
>> different md5sum as it assumes tv for in and pc for out so
>> stretches to full range.
>
> So how can we reproduce the bug you experienced?

Well, I am not sure I would call it a bug as that implies some intended
behavior and I have no idea what yuvj was ever supposed to mean to
ffmpeg - or indeed what yuv means WRT subsampling positions.

The test just answered my own curiosity/speculation as to whether using
target yuvj caused swscaler to do more than just mess with levels if
subsampling was involved.

the test went something like

ffmpeg -i ramp-yuv444p.nut -pix_fmt yuv420p -vf
scale=in_range=pc:out_range=pc -vcodec rawvideo yuv420p-pc-pc.nut

repeat for tv pc and pc tv (tv tv is the same as pc pc = do nothing to
levels)

Then do it all again with -pix_fmt yuvj420p and see that the md5sums are
the same and conclude there is no difference in sub-sampling with target
yuv420p and yuvj420p.

Of course tests can always be better - maybe a single frame 7 band ramp
is not the best input - but then I would have thought there would have
been some diff. 7 bands = 7 ramps of different colours not step size.



More information about the ffmpeg-user mailing list