[FFmpeg-devel] [PATCH] avfilter: add audio surround upmixer

[Paul B Mahol]

On 5/26/17, Carl Eugen Hoyos <ceffmpeg at gmail.com> wrote:
> 2017-05-25 22:39 GMT+02:00 Paul B Mahol <onemda at gmail.com>:
>> On 5/25/17, Carl Eugen Hoyos <ceffmpeg at gmail.com> wrote:
>>> 2017-05-25 16:45 GMT+02:00 Paul B Mahol <onemda at gmail.com>:
>>>
>>>> + at section surround
>>>> +Apply audio surround upmix filter.
>>>> +
>>>> +This filter allows to produce multichannel output from
>>>> stereo audio stream.
>>>
>>> Does this work on Pro Logic-encoded files (ticket #4085)
>>> or does this filter use a different algorithm?
>>
>> It works with sample from that ticket.
>
>> It uses own algorithm.
>
> Is it related to one of the algorithms listed here?
> https://en.wikipedia.org/wiki/Matrix_decoder
> It may make sense to mention the used algorithm in the
> documentation (or the source code).

All those are passive algorithms, while this one is active.

>
>>> Does a filter have advantages over an implementation in
>>> libswresample?
>>
>> libswresample doesnt do this.
>
> Yes.
> My question was if your implementation shouldn't be
> part of libswresample, or rather what would be the
> disadvantage.

This filter relies on FFT, and it should probably stay in libavfilter library.

>
>>> Iirc, our Pro Logic downmixer has an issue with "phase shift"
>>> (ticket 3455 / 4175): Does this filter give expected output
>>> for FFmpeg-encoded files or original Pro Logic files (or is
>>> there no difference on upmixing)?
>>
>> FFmpeg downmixing, while certainly not correct, is giving
>> reasonable output when used with this filter.
>
> But it works better for Pro Logic files?

I only have one such file, and its not that useful.

FFmpeg have only downmixing from 5.1 layout to stereo.

I also plan to add 5.1 to 7.1 upmixing.