[FFmpeg-devel] [PATCH] Change libvpx-vpx default to crf=32.

[Elliott Karpilovsky]

On Thu, Aug 29, 2019 at 10:14 AM Jan Ekström <jeebjp at gmail.com> wrote:
>
> On Thu, Aug 29, 2019 at 7:17 PM Carl Eugen Hoyos <ceffmpeg at gmail.com> wrote:
> >
> > Am Do., 29. Aug. 2019 um 18:12 Uhr schrieb Elliott Karpilovsky
> > <elliottk-at-google.com at ffmpeg.org>:
> > >
> > > On Wed, Aug 28, 2019 at 3:11 PM Carl Eugen Hoyos <ceffmpeg at gmail.com> wrote:
> > > >
> > > > Am Mi., 28. Aug. 2019 um 23:26 Uhr schrieb Elliott Karpilovsky
> > > > <elliottk-at-google.com at ffmpeg.org>:
> > > > >
> > > > > Current default is 200kbps, which produces inconsistent
> > > > > results (too high for low-res, too low for hi-res). Use
> > > > > CRF instead, which will adapt. Affects vp8/vp9. Also
> > > > > have VP8 use a default bitrate of 256kbps.
> > > >
> > > > Would it simplify the patch if -crf 0 would indicate lossless encoding?
> >
> > > I do not believe so. I believe that there would be more code changes
> > > to switch the checks from ctx->lossless to ctx->crf,
> >
> > Of course but that wasn't my question.
> > I believe it is a horrible bug that x264 uses crf 0 for lossless encoding,
> > while libvpx uses another option. If fixing this bug simplifies your
> > patch, I believe it is worth the effort.
> >
>
> Just for the record, libx264 only does lossless for CRF=0 for 8bit
> coding. The quantizer range for CRF was expanded downwards for higher
> bit depths since the results were similar for lossy coding. In that
> case you either have to poke the right negative value, or set
> quantizer to zero - which is the option that works for all bit depths
> with libx264.
>
> That said, I do not have heavy objections one way or another regarding
> how the libvpx wrapper defines CRF (as I think the definition for VP8
> and VP9 is already in general different from what libx264 does - like
> bit rate being required for VP8 etc). Just wanted to refresh people's
> minds regarding libx264's CRF behavior with lossless coding.
>
> Best regards,
> Jan

Thanks for the context. My preference is to keep this patch about
changing the default behavior. I would not oppose someone else writing
another patch to change CRF=0 to lossless, but it feels like these two
should be separate.