Where is the new super codec? (the Alliance for Open Media) ?

[Stears555]

Hi!

I want to install it on my computers.....
Where can I found it?

According to advanced codecs the new super codec? (the Alliance for Open Media) was released....

https://www.videohelp.com/software/Win7codecs/version-history#changelog

- update LAV filters 68.1.35 (un-official release)
- update Gabest filters 1.5.2172 (official release)
- update MediaInfo DLL 7.90 (official release)
- update Icaros 3.01RC1 (official release)
- update support for the Alliance for Open Media (AV1 video playback and thumbnails)
- fully support 3D playback - see Setup and usage for more info

The name of the new codec is AOM1.... (AOMEdia Video1)
https://en.wikipedia.org/wiki/AOMedia_Video_1

[sneaker]

AOMedia AV1 isn't yet finalized. That's just experimental support that may break with the next encoder update. (And please don't use ridiculous names like "super codec" based on their marketing claims.)

[Stears555]

AOMedia AV1 isn't yet finalized. That's just experimental support that may break with the next encoder update. (And please don't use ridiculous names like "super codec" based on their marketing claims.)

HEVC was a great disappointment , because HEVC was not a real break-through like the AVC was in its beginning...

[sneaker]

Then I predict you will be disappointed by AV1 as well.

[godai]

what its wrong with hevc, im happy with that one, compression its amazing. imo.

[pandy]

Where can I found it?

Check attachment.

[KarMa]

Any kind of documentation to go with that pandy? Can't seem to figure out how to do anything more than the default. Don't know if it has presets (fast, medium, slow) or how to change specific encoding settings. Anyway nice share, don't know where you got it.

Update: Ok wow I'm an idiot ignore me.

[hello_hello]

what its wrong with hevc, im happy with that one, compression its amazing. imo.

For standard definition at least, it's not exciting me too much yet.
Unfortunately I don't think hardware player support for 10 bit h264 decoding will ever become mainstream, whereas 10 bit h265 decoding no doubt will (for h265 10 bit was included in the spec from day one), and that'll probably be a factor eventually, even for lower resolutions. x265 will probably come into it's own for UHD and x264 won't compete there.

I was playing around with some noise filtering today, so while I was at it I re-encoded a small section as a lossless file (after noise filtering) and ran some test encodes. I probably should have made it a bit longer, but the lossless source was 2000 frames at 1024x576.

The medium speed preset and CRF18 for each encode
x265 = 3540 kbps
x264 = 4960 kbps
x264, Tune Film = 5166 kbps

x265 vs x264 at CRF18 - x265 blurs more detail (sometimes it's a though it has a noise filter) and it's more inclined to "warp" the shape of objects, but the bitrate was a fair bit lower, so I ran one more x265 encode using 2 passes to specify a bitrate of 4960. That improved the quality, but I still thought the x264 encodes looked more accurate and retained more detail. It's probably going to be nitpicking for some people because to spot differences reliably I had to run the 1024x576 encodes fullscreen on my TV, pause each on the same frame and switch between them. If I did that at a 1024x576 resolution, the video would only fill about one quarter of the screen and I'd struggle to spot any differences. Still, the quality is either the same or it's not even if you're less likely to notice differences while the pictures are moving. That's how these screenshots were taken.... while the video was being upscaled to 1080p by MPC-HC on my TV (Bicubic upsizing). In that shot, the camera is moving to the right as the guy pushes the bike so nothing is static, but the motion isn't exactly hectic.

The lossless source:


x264, CRF18, Tune Film:


X265, 2 pass, 4960 kbps:


My x265 critique compared to x264:
The grass is a little blurrier.
Well everything is a little blurrier but it's noticeable looking at the front wheel of the bike, and the gravel behind the wheel has taken some extra blurring compared to the gravel elsewhere.
It seems somebody washed the freeway pillars for the x265 encode. What may appear to be detail when you look at the pillars in the source is largely the remaining noise, but still, x265 blurred it.
There's less detail in the dirt on the side of the car.
Behind the guy's elbow there's a pillar and behind that plants on the slope. Where the pillar disappears behind the guy's arm a small section of the plant behind the pillar appears to be in front of it.

I'm very likely to not get a chance till the weekend now, but I intend to repeat those encodes at 720p and 1080p to see how things change as the resolution increases (the source was 1080p before I de-noised and downscaled).

[KarMa]

@hello_hello
x265 supports --tune grain, may want to give that a try. It can look pretty good but basically doubles encode times for the same preset with no tuning.

[poisondeathray]

Nightly builds can be found here

http://tmod.nmm-hd.org/aom/

-help for instructions / switches

[vhelp]

H265 / HEVC is a flop, a failure, in my opinion.

In the clips you posted, you can see the level of detail that x264 retained, while x265 continues to apply some form of noise reduction to keep the filesize lower in some cases.

x265 will never look at good as x264. I'm convinced. And after so many test encodes of all sorts of clips, I just can't get past how much x265 "smooths" out the detail out of the video. And --tune grain does nothing to help but worsens it in some if not all cases. But i'm still testing various scenarios. There must be a setting or group of settings that need to either be excluded or finely tuned to *stop* the noise reduction that x265 is applying (in order to reduce filesize) to the video. Otherwise, trying to match x265 to x264 is a no-win scenario and continuing on this gurney is like beating a dead horse, to coin the phrase.

h265, via x265.exe, as an advanced codec that was suppose to super-seed (spelling) x264 has, in my opinion, failed. All encoders available for testing this has proven that. AMD, NVidia, and Intel (quicksync) and x265 has all proven this format's failure.

[vhelp]

The only way H265 / HEVC via x265 will compete with x264 (H264/AVC) is if the x265 team copy the logic/algorithm that x264 uses to retain detail. In my opinion, the "retaining the detail" part was not part of the h264/avc spec (was never realized) and a crafty method was invented by the developer, thanks to the many testers (doom9) that revealed this important fact during the x264 encoder development stages over the years.

[pandy]

Anyway nice share, don't know where you got it.

Compiled with help of https://github.com/jb-alvarado/media-autobuild_suite - great suite as you can enable nonfree option and as such use codecs normally not available due license restrictions.

[hello_hello]

@hello_hello
x265 supports --tune grain, may want to give that a try. It can look pretty good but basically doubles encode times for the same preset with no tuning.

Next time I run some tests I'll give it a try, but it's already slower than x264 and not managing the same quality at a lower bitrate (at least not for standard definition), so even if --tune grain does the trick it'll probably end up similar quality to x264, only much slower.

vhelp,
x264 has some psycho-visual options. They're not the same as for x264 with different names?
http://x265.readthedocs.io/en/default/cli.html#psycho-visual-options

You did cause me to remember though.... back in the earlier x264 days, before the psy stuff was introduced, I recall the main x264 complaint was in regard to it's burring of detail. I wasn't using x264 myself back then, but I do recall it was fairly standard to reduce the deblocking as much as possible to retain maximum detail. Something like deblock=1:-3:-3 would have been commonly used. Next time I play with x265 I'll try a similar thing to see if it helps (unless someone knows I shouldn't bother) and it appears for h265 it's possible to disable de-blocking entirely.

[vhelp]

After some time has passed and tweaking around, I may have to take back what I said in post # 11 and # 12 for x265 encoded videos. The encoder appears to keep the detail provided you used certain parameters wisely. I'm running some more test encodes and will eventually change my mind once I'm satisfied for sure. However, as for QSV. I don't know. I can bring out the details also, but the filesize will be twice or more times larger than if encoded by x264 and x265, but faster encoding times. I'm still performing various encoding tests on my Kabylake QSV and will eventually report back about that, possibly in another thread.

I do not like to encode my videos to a low bitrate. I have now passed that stage and practice. I believe that a video should be encoded to show no artifacts, as best as possible, and not try to show off how low a bitrate we can make a given video, whether to show off an argument about any encoder's capability or other reason. I have ripped a number of Hollywood (bluray) videos and see that they are mainly 25 Mbps, but I am still seeing members here posting examples down to 4 Mbps, which is not a good thing. And, although the video may look fine on a small screen, or laptop screen, the truth comes out when you view a much larger screen. When I encode videos at 4 Mbps, I see a big difference between my 15" laptop monitor versus my 22" monitor, where both have different Gamma and Contrast setup. I mean. Wow! What was I thinking? 4 Mbps is way too low. The regret will set in once you come to terms with it. Anyway. I think that 9 to 15 Mbps (or higher) is a better bitrate strategy, especially since the larger monitors and tv's will ultimately be used. In fact, here is my suggestion. test and determine several short encoded videos at 9, 15, and 25 Mbps bitrate and see where you finally see clean, undisturbed images. No artifacts, no jibblets, no sectional or matrice type flutter or shimmering, etc etc. Then that will finalize your ultimate bitrate for that given video. This is for those videos that you want absolute highest quality. If you just want to encode tv shows and other such that don't have good quality to begin with, then use your best judgement and use what works best in your view and learn to live with the outcome of those videos moving forward. But eventually, you may come to a point in your viewing pleasures and change even that, to something better.

EDIT: corrected KBps to Mbps

[sneaker]

"KBps"?

[sophisticles]

I think that many of the complaints people are posting here of x265 (and HEVC in general) has to do with using presets for encoding. In my tests, the x265 defaults (especially when used via handbrake for some reason) tend to blur/smooth out a lot of details, no matter how much bit rate you throw at the encode.

BUT, if you use Hybrid (for easy configuration) and 12 bit 4:4:4 x265 and play around with psy-rdoq, you can get some very impressive quality encodes, which I can't replicate with x264 no matter how aggressive a setting I use.

[hello_hello]

12 bit, 4:4:4, x265 and fiddling with psy-rdoq vs x264 set aggressively. It doesn't mean much.

I think part of the problem is the people posting here are testing the types of sources they'd typically encode, using settings that are well supported by hardware players today or likely to be widely supported in the near future, and I'm fairly sure encoding with Handbrake would be far closer to that than a 4:4:4 source encoded with h265 12 bit.

When I tested with a typical 4:2:0 standard definition source, x265 10 bit encodes with the default settings still blurred more than x264 8 bit encodes with Tune Film at the same bitrate, and there was very little difference between 8 and 10 bit x265 in respect to blurring. That's probably going to be the consumer level limit for the foreseeable future. h264 will probably be limited to 8 bit and 4:2:0 forever if you want your videos to be playable on devices other than a PC, while hardware player support for 10 bit h265 4:2:0 will hopefully become mainstream some time in the future. Will 12 bit h265 support by consumer players become common any time in the foreseeable future, and if so is it likely to include anything other than 4:2:0?

[Cornucopia]

Funny how "cool new super codec" has devolved into x264 vs x265 and tuning & colorspace tweaks.

I would be complaining about how it has gotten so far off track from the OT, but honestly it is a vast improvement. At least now there is helpful, actionable tips.

Scott