[FFmpeg-devel] [PATCH] lavdevice: Add VideoToolbox output device.

[Thilo Borgmann]

Hi,

Am 08.06.20 um 16:27 schrieb Nicolas George:
> Thilo Borgmann (12020-06-08):
>> How do you avoid the auto-scaler to be scaling to alsa's default
>> format of pcm_s16 on the command line?
> 
> We cannot, not with your original version, not with how alsa_enc does
> and the new version.
> 
> Because they are not sample formats, they are codecs, and converting
> from a codec to another is more complex than just a resampler. And even
> if in these particular cases they are trivial codecs, they could easily
> not be: ALSA supports S/PDIF, so it could be AC3 or something.

yes I figured that codec thing out after querying on IRC. So with providing -a:c XY I can do all the formats (or pcm codecs so to speak) in one muxer where I had seperate muxers for before. So v2 works like the examples you gave for alsa. Might even work for compressed codecs depending on what OSX accepts.

However you say we cannot avoid the auto scaler? I'm not sure what you mean we cannot do...
And do you think v2 is still object for improvements regarding this?


> The solution to have format negotiation would be to make sure all output
> devices have a matching lavfi sink. Unfortunately, for now, I am not
> sure lavfi's scheduling work with non-buffer sinks. And it has other
> drawbacks, including preventing S/PDIF.

This let's me think you're referring to some more general design shortcomings. Not sure if you relate that to the patch or something in general we should do prior to do the audiotoolbox output device.

Thanks,
Thilo