libvpx 1.5.0 release soon

[Johann Koenig]

It's been about 6 months since the 1.4.0 release and libvpx has seen
substantial improvements to vp9 encoding speed and quality, more than
a few changes to vp8 to improve WebRTC performance, and massive code
movement and cleanup.

We would like to pick a release candidate within the next few weeks.
If you have issues of any kind with current HEAD please reply or file
a bug[0]. If you have any ideas for improvements please file a bug as
well and we can attempt to fix it for 1.5.0 or triage it for a future
release.

[0] http://www.webmproject.org/code/bug-reporting/

[David Pio]

awesome. do you have a sense for a % speedup in vp9 encoding speed since 1.4.0? or maybe a 'top hits' for tools and features within vp9 that have improved in perf?

Thanks!

pio

--
You received this message because you are subscribed to the Google Groups "Codec Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to codec-devel...@webmproject.org.
To post to this group, send email to codec...@webmproject.org.
Visit this group at http://groups.google.com/a/webmproject.org/group/codec-devel/.
For more options, visit https://groups.google.com/a/webmproject.org/d/optout.

[Michael IV]

What about alpha transparency support? 

[Johann Koenig]

On Wed, Sep 30, 2015 at 10:22 AM, David Pio <dpi...@gmail.com> wrote:
> awesome. do you have a sense for a % speedup in vp9 encoding speed since
> 1.4.0? or maybe a 'top hits' for tools and features within vp9 that have
> improved in perf?

Unfortunately I don't have anything specific to call out at the
moment. We'll run numbers and have some more information as the
release is finalized.

[Johann Koenig]

On Wed, Sep 30, 2015 at 10:27 AM, Michael IV <explo...@gmail.com> wrote:
> What about alpha transparency support?

Alpha was never part of the VP9 bitstream. It's currently in the same
position as VP8 - it is possible to add alpha information outside the
bitstream:
https://developers.google.com/web/updates/2013/07/Alpha-transparency-in-Chrome-video?hl=en

However I'm not aware of any recent work on that effort.

[Sagara Wijetunga]

Gave it try on chroma 422->420, 420->420 up to 1080p. Seems alright. Noticeably faster, probably 20%+.

I used ffmpeg 2.8 built with libvpx git on Mac OS X.

The -speed parameter does not seems to work in pass 2, does it only work in pass 1?


Regards
Sagara

[Johann Koenig]

I'm not actually sure what -speed does in ffmpeg or how it's hooked
up. It's not mentioned in any of the vp8 documentation or general
info. There is mention of -cpu-used which is the vpxenc way of
adjusting speed. 0 is very slow, best compression, 16 is the fastest
setting (but I think only up to about 8 or 10 are defined for vp9):
http://wiki.webmproject.org/ffmpeg
https://trac.ffmpeg.org/wiki/Encode/VP8

There are preset files mentioned:
http://ffmpeg.org/ffmpeg.html#ffpreset-files

-speed is mentioned on the vp9 page but the whole page only has 2 pass examples:
http://wiki.webmproject.org/ffmpeg/vp9-encoding-guide

Perhaps someone more familiar with ffmpeg could comment.

[James Zern]

-speed is an alias for -cpu-used left around for compatibility. Setting it in 2-pass to 0/1/2 should have noticeable effect on both encode time and quality as mentioned above.

http://ffmpeg.org/ffmpeg-codecs.html#libvpx

 

[Sagara Wijetunga]

Hi James, thanks for the info. I found the issue, -cpu-used was already there in the .ffpreset file set to 0. That's why my -speed was not effective.

So now the -speed is working and the -speed 1 is about 10 times faster the -speed 0 for 480p @ crf 33.

Looking forward to see the 1.5.0 soon.

Sagara