[FFmpeg-user] FFv1: newer ffmpeg builds create larger files?

[Paul B Mahol]

On 6/18/12, Peter B. <pb at das-werkstatt.com> wrote:
> Quoting Michael Niedermayer <michaelni at gmx.at>:
>
>> On Mon, Jun 04, 2012 at 07:31:34PM +0200, Peter B. wrote:
>>> On 06/04/2012 02:56 AM, Michael Niedermayer wrote:
>>> > On Mon, Jun 04, 2012 at 01:22:27AM +0200, Peter B. wrote:
>>> >> Hm... I don't see any information about the bitstream parameters.
>>> >> I've tried 2 things with "-debug 1":
>>> >> 1) calling ffprobe on the original
>>> >> 2) transcoding the original as before
>>> > [...]
>>> >> Is there something I've misunderstood/overlooked?
>>> > ffmpeg -debug 1 -i ffv1file -f null -
>>> >
>>> Thank you! Interesting commandline :)
>>>
>>> Interestingly, this showed "coder:0" for the original (=smallest) file,
>>> whereas using "-coder 0" resulted in larger files than "-coder 1" in my
>>> tests.
>>>
>>> When I checked files encoded with "-coder 1", the above command output
>>> "coder:2" - I've verified this behavior, and all files encoded with
>>> "-coder 1" are being shown with "-coder 2". Why is that?
>>
>> The difference between coder type 1 and 2 are documented in the spec
>> coder type 2 performs better than 1 in all cases i remember
>> so its used by default when range coding is requested
>
> I think there's a misunderstanding here:
> I thought that the valid values for "-coder" are "0" and "1", not "1"
> and "2". Correct me if I'm wrong.
>
> Secondly, the puzzling part is, that the ffv1-files I'm getting from
> the realtime encoding during capture (using ffdshow-tryouts with ffv1
> version 1), are way smaller than *any* ffv1-encoded version of the
> same content, transcoded with the current git-version(s) of ffmpeg.
>
> Why is it that I'm unable to reproduce these small files - regardless
> which parameters I use?

Are you sure that pixel formats are same in both files?