[FFmpeg-devel] [PATCH][TESTERS WANTED] avfilter: add apitch filter

[Paul B Mahol]

Hi,

On 5/12/19, Nicolas George <george at nsup.org> wrote:
> Paul B Mahol (12019-05-11):
>> It is currently separate filter for ease of testing.
>
> Then no problem. But I insist: once it is in shape, for the convenience
> of the users and ease of maintenance, it should be a single filter.
>
> Of course, if you disagree, we can discuss it here.
>
>> It can not be "part" of current atempo implementation for numerous
>> reasons,
>> most significant one being bad output quality with < 1 tempo scale
>> factor of current atempo implementation.
>
> This is one of the reasons it should be a single filter: the users
> should not need to know the arbitrary limits of different filters
> implementation. The filters are there to perform a task, and they should
> select the best implementation for it automatically.

That is hard problem, how would one know what implementation is best for
user scenario.

>
>> Also current atempo implementation have limits for minimal scale
>> reduction,
>
> IIRC, the limits were rather arbitrary. But even if they are not, my
> point stand: the filter should select the best implementation for the
> requested parameters, automatically without bothering the users.

I disagree, current filter have artifacts with small scale factors.

>
>> unlike rubberband filter in libavfilter. Also rubberband filter having
>> similar and bigger functionality.
>
> Oh, great, another filter with the same task. If I had noticed at the
> time, I would have insisted that it be part of atempo too at the time.

Looks like I missed to explain it in more fashion to you full understanding,
it is filter that uses external library, do you still insist it to be
part of atempo?

>
>> That's way I request for quality testers.
>> Best would be to compare this filter with rubberband and current atempo
>> filter.
>>
>> This filter is supposed to be current atempo full replacement.
>> (atempo would be just this filter but with same first two options but
>> with swapped positions).
>
> Swapping the position will break existing scripts. It should be avoided.

Looks like you missed to understand my detailed explanation.
Lets try it to explain in again, in more detailed fashion:
I mean there would be two filters doing same thing:
apitch with 1st option pitch and 2nd option tempo.
atempo with 1st option tempo and 2nd option pitch.
These filters would share same C file.

Regards, have a nice day