[Cin] Libaom 3.4.0

Tue Jun 28 20:53:23 CEST 2022

I did not receive your e-mail in my mailbox. (I have sick people at my providers', sometimes they deliberately screw things.)
So I went to see, and I found you replied to me:
https://lists.cinelerra-gg.org/pipermail/cin/2022-June/005284.html

And this mail is reconstruted from the web page, I got the "In-Reply-To:" string from your mail by clicking on the very first link, the one at the top of the page from your mail, so this should look like a correct reply with all the right headers. Unfortunately clicking on that link and using Mutt to write a temporary mail did not get me the contents of your email, so I'll simply paste it in here from that wab page in my normal working Mutt.

Apologies for this inconvenience.

> вт, 28 июн. 2022 г., 16:41 Miroslav Rovis <miro.rovis at croatiafidelis.hr>:
> 
> > On 220627-10:03+0200, Andrea paz via Cin wrote:
> > > About AV1 and ffmpeg there is also this commit:
> > >
> > https://github.com/FFmpeg/FFmpeg/commit/a44fba0b5b3b4090f9238751736198ddd1f0f1d5
> > > For now, AV1 is not a convenient codec for me, given its poor
> > > performance in both decoding and encoding.
> > > --
> > > Cin mailing list
> > > Cin at lists.cinelerra-gg.org
> > > https://lists.cinelerra-gg.org/mailman/listinfo/cin
> > And the thread started when developer Andrew gave:
> > https://groups.google.com/a/webmproject.org/g/webm-discuss/c/D5kHbR8-GaU
> > which indeed has a lot of AV1 (case insensitive).
> > And it is AV1:
> > https://trac.ffmpeg.org/wiki/Encode/AV1
> >
> > I've recently had huge losses of quality with VP9 encoding directly in
> > Cinerella, via FFmpeg of course. Once I reverted to encode to MP4, quality
> > was fine again, and the size compared to original, smaller (well, certainly
> > not so small as AV1, far from, though).
> >
> > So I'm afraid, VP9 is unacceptably poor perfomance. (Or I have not been
> > able to set the right options, although I tweaked quite a lot; I do want
> > good compression, and I found no such options to retain quality while
> > giving not very large WEBM file to publish online... To the contrary, AV1
> > can really really do that!)
> >
> 
> Well, you tried two-pass encoding? I mostly encode screencasts and there
> vp9 as presented by Cingg presets works reasonably well for my limited
> eyesight ..
My eyesight is poor too, and getting poorer. How I compare the quality btwn the original and the worked video is comparing frame by frame. Running mplayer <the-original> from one terminal, and mplayer <the-worked-video> from another termianl. Get it to same position, and then advance frame by frame, which is '.' key in mplayer. Very tedious...
Not tried 2-pass yet in Cinelerra, I did 2-pass in plain FFmpeg. I probably should try it in Cinelerra too.

But I saw in the ML that you've relatively recently pushed lossless VP9 encoding, I'm yet to find if it is available in my Cinelerra that I built from git yesterday (and which I can't find the version of in the About page, but I got the git very little time before I built it. I built it from master branch. The tag in git is 2022-05.)

> >
> > I'm familiar enough with what AV1 codec practically entails. It's been
> > huge computing power needed for even short encodings. Such as, the shortest
> > of the videos (27s):
> >
> > https://croatiafidelis.hr/2021/rpi-CM4/VID_20210213_150554_CM4IO_only_on_PSU_voltage_measurement_RED_LED_and_BEEPS.webm
> > on https://croatiafidelis.hr/2021/rpi-CM4/210213-150554.php (not all
> > players/browsers can play it)
> > It's the future probably.
> >
> > I only have:
> > # cat /proc/cpuinfo
> > [...]
> > model name      : AMD Ryzen 7 PRO 4750G with Radeon Graphics
> > [...]
> > which is modest but is not a feeble processor.
> >
> > Do these changes make any really significant speedup for encoding AV1,
> > will there be any significanlty smaller power requirements for encoding
> > AV1, with FFmpeg (in Cinelerra and otherwise)? Or will it be practically
> > off the table with non-top-powerful processors like mine, for any footage
> > that's not just very short clips?
> >
> 
> Well, speed difference said to be in 10-20 percent range relative to
> previous libaom release ...so ...nothing groundbreaking?
> 
> There are supposedly faster encoders (ra1ve and svt-av1 ? ) but their
> quality said to be less than libaom's ...
May be I could try some of those too...

> VDPAU is decoing-only interface, so for encoding you probably need very new
> and costly intel hardware and fresh libva ..
> 
> https://www.phoronix.com/scan.php?page=news_item&px=VA-API-libva-2.14-Released

Thanks for the information.
> 
> No sign of av1 encoding support in mesa (for AMD hardware)  so far ....
So far... I'm sure they're catching up in secret. But that won't be cheap either.
Or maybe not:
https://en.wikipedia.org/wiki/AV1#Hardware
and:
https://www.reddit.com/r/AV1/comments/c7aaj5/av1_encoder_with_amd_app_support/

Bleak prospect for me. I have two AM4-socket boards based systems, can only stay with AMD now, as I am rather poor.

Regards!

--
Miroslav Rovis
Zagreb, Croatia
https://www.CroatiaFidelis.hr
my PGP-key:
https://www.croatiafidelis.hr/FCF13245ED247DCE443855B7EA9884884FBAF0AE.asc
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.cinelerra-gg.org/pipermail/cin/attachments/20220628/7a037e94/attachment.asc>