NHacker Next
login
▲AV1: A Modern, Open Codecnetflixtechblog.com
183 points by CharlesW 3 hours ago | 82 comments
Loading comments...
notatoad 5 minutes ago [-]
I understand that sometimes the HN titles get edited to be less descriptive and more generic in order to match the actual article title.

What’s the logic with changing the title here from the actual article title it was originally submitted with “AV1 — Now Powering 30% of Netflix Streaming” to the generic and not at all representative title it currently has “AV1: a modern open codec”? That is neither the article title nor representative of the article content.

crazygringo 2 hours ago [-]
Wow. To me, the big news here is that ~30% of devices now support AV1 hardware decoding. The article lists a bunch of examples of devices that have gained it in the past few years. I had no idea it was getting that popular -- fantastic news!

So now that h.264, h.265, and AV1 seem to be the three major codecs with hardware support, I wonder what will be the next one?

0manrho 29 minutes ago [-]
> To me, the big news here is that ~30% of devices now support AV1 hardware decoding

Where did it say that?

> AV1 powers approximately 30% of all Netflix viewing

Is admittedly a bit non-specific, it could be interpreted as 30% of users or 30% of hours-of-video-streamed, which are very different metrics. If 5% of your users are using AV1, but that 5% watches far above the average, you can have a minority userbase with an outsized representation in hours viewed.

I'm not saying that's the case, just giving an example of how it doesn't necessarily translate to 30% of devices using Netflix supporting AV1.

Also, the blog post identifies that there is an effective/efficient software decoder, which allows people without hardware acceleration to still view AV1 media in some cases (the case they defined was Android based phones). So that kinda complicates what "X% of devices support AV1 playback," as it doesn't necessarily mean they have hardware decoding.

JoshTriplett 2 hours ago [-]
> So now that h.264, h.265, and AV1 seem to be the three major codecs with hardware support, I wonder what will be the next one?

Hopefully AV2.

jsheard 2 hours ago [-]
H266/VVC has a five year head-start over AV2, so probably that first unless hardware vendors decide to skip it entirely. The final AV2 spec is due this year, so any day now, but it'll take a while to make it's way into hardware.
adgjlsfhk1 2 hours ago [-]
H266 is getting fully skipped (except possibly by Apple). The licensing is even worse than H265, the gains are smaller, and Google+Netflix have basically guaranteed that they won't use it (in favor of AV1 and AV2 when ready).
2 hours ago [-]
adzm 2 hours ago [-]
VVC is pretty much a dead end at this point. Hardly anyone is using it; it's benefits over AV1 are extremely minimal and no one wants the royalty headache. Basically everyone learned their lesson with HEVC.
kevincox 2 hours ago [-]
If it has a five year start and we've seen almost zero hardware shipping that is a pretty bad sign.

IIRC AV1 decoding hardware started shipping within a year of the bitstream being finalized. (Encoding took quite a bit longer but that is pretty reasonable)

jsheard 1 hours ago [-]
https://en.wikipedia.org/wiki/Versatile_Video_Coding#Hardwar...

Yeah, that's... sparse uptake. A few smart TV SOCs have it, but aside from Intel it seems that none of the major computer or mobile vendors are bothering. AV2 next it is then!

shmerl 33 minutes ago [-]
When even H.265 is being dropped by the likes of Dell, adoption of H.266 will be even worse making it basically DOA for anything promising. It's plagued by the same problems H.265 is.
thrdbndndn 1 hours ago [-]
I'm not too surprised. It's similar to the metric that "XX% of Internet is on IPv6" -- it's almost entirely driven by mobile devices, specifically phones. As soon as both mainstream Android and iPhones support it, the adoption of AV1 should be very 'easy'.

(And yes, even for something like Netflix lots of people consume it with phones.)

dylan604 2 hours ago [-]
how does that mean "~30% of devices now support AV1 hardware encoding"? I'm guessing you meant hardware decoding???
crazygringo 2 hours ago [-]
Whoops, thanks. Fixed.
dehrmann 2 hours ago [-]
Not trolling, but I'd bet something that's augmented with generative AI. Not to the level of describing scenes with words, but context-aware interpolation.
km3r 2 hours ago [-]
https://blogs.nvidia.com/blog/rtx-video-super-resolution/

We already have some of the stepping stones for this. But honestly much better for upscaling poor quality streams vs just gives things a weird feeling when it is a better quality stream.

randall 2 hours ago [-]
for sure. macroblock hinting seems like a good place for research.
snvzz 2 hours ago [-]
>So now that h.264, h.265, and AV1 seem to be the three major codecs with hardware support

That'd be h264 (associated patents expired in most of the world), vp9 and av1.

h265 aka HEVC is less common due to dodgy, abusive licensing. Some vendors even disable it with drivers despite hardware support because it is nothing but legal trouble.

IgorPartola 2 hours ago [-]
Amazing. Proprietary video codecs need to not be the default and this is huge validation for AV1 as a production-ready codec.
pbw 2 hours ago [-]
There's an HDR war brewing on TikTok and other social apps. A fraction of posts that use HDR are just massively brighter than the rest; the whole video shines like a flashlight. The apps are eventually going to have to detect HDR abuse.
thrdbndndn 1 hours ago [-]
The whole HDR scene still feels like a mess.

I know how bad the support for HDR is on computers (particularly Windows and cheap monitors), so I avoid consuming HDR content on them.

But I just purchased a new iPhone 17 Pro, and I was very surprised at how these HDR videos on social media still look like shit on apps like Instagram.

And even worse, the HDR video I shoot with my iPhone looks like shit even when playing it back on the same phone! After a few trials I had to just turn it off in the Camera app.

johncolanduoni 20 minutes ago [-]
I wonder if it fundamentally only really makes sense for film, video games, etc. where a person will actually tune the range per scene. Plus, only when played on half decent monitors that don’t just squash BT.2020 so they can say HDR on the brochure.
Forgeties79 34 minutes ago [-]
The only time I shoot HDR on anything is because I plan on crushing the shadows/raising highlights after the fact. S curves all the way. Get all the dynamic range you can and then dial in the look. Otherwise it just looks like a flat washed out mess most of the time
munificent 2 hours ago [-]
Just what we need, a new loudness war, but for our eyeballs.

https://en.wikipedia.org/wiki/Loudness_war

eru 2 hours ago [-]
Interestingly, the loudness war was essentially fixed by the streaming services. They were in a similar situation as Tik Tok is now.
crazygringo 2 hours ago [-]
This is one of the reasons I don't like HDR support "by default".

HDR is meant to be so much more intense, it should really be limited to things like immersive full-screen long-form-ish content. It's for movies, TV shows, etc.

It's not what I want for non-immersive videos you scroll through, ads, etc. I'd be happy if it were disabled by the OS whenever not in full screen mode. Unless you're building a video editor or something.

JoshTriplett 2 hours ago [-]
Or a photo viewer, which isn't necessarily running in fullscreen.
jsheard 2 hours ago [-]
Sounds like they need something akin to audio volume normalization but for video. You can go bright, but only in moderation, otherwise your whole video gets dimmed down until the average is reasonable.
sieabahlpark 2 hours ago [-]
[dead]
recursive 2 hours ago [-]
My phone has this cool feature where it doesn't support HDR.
JoshTriplett 2 hours ago [-]
That's true on the web, as well; HDR images on web pages have this problem.

It's not obvious whether there's any automated way to reliably detect the difference between "use of HDR" and "abuse of HDR". But you could probably catch the most egregious cases, like "every single pixel in the video has brightness above 80%".

kmeisthax 54 minutes ago [-]
Funnily enough HDR already has to detect this problem, because most HDR monitors literally do not have the power circuitry or cooling to deliver a complete white screen at maximum brightness.

My idea is: for each frame, grayscale the image, then count what percentage of the screen is above the standard white level. If more than 20% of the image is >SDR white level, then tone-map the whole video to the SDR white point.

JoshTriplett 14 minutes ago [-]
That needs a temporal component as well: games and videos often use HDR for sudden short-lived brightness.
eru 2 hours ago [-]
> It's not obvious whether there's any automated way to reliably detect the difference between "use of HDR" and "abuse of HDR".

That sounds like a job our new AI overlords could probably handle. (But that might be overkill.)

ElasticBottle 2 hours ago [-]
Can someone explain what the war is about?

Like HDR abuse makes it sound bad, because the video is bright? Wouldn't that just hurt the person posting it since I'd skip over a bright video?

Sorry if I'm phrasing this all wrong, don't really use TikTok

JoshTriplett 2 hours ago [-]
> Wouldn't that just hurt the person posting it since I'd skip over a bright video?

Sure, in the same way that advertising should never work since people would just skip over a banner ad. In an ideal world, everyone would uniformly go "nope"; in our world, it's very much analogous to the https://en.wikipedia.org/wiki/Loudness_war .

johncolanduoni 17 minutes ago [-]
Not everything that glitters (or blinds) is gold.
dylan604 2 hours ago [-]
sounds like every fad that came before it where it was over used by all of the people copying with no understanding of what it is or why. remember all of the HDR still images that pushed everything to look post-apocalyptic? remember all of the people pushing washed out videos because they didn't know how to grade the images recorded in log and it became a "thing"?

eventually, it'll wear itself out just like every other over use of the new

kmeisthax 59 minutes ago [-]
I would love to know who the hell thought adding "brighter than white" range to HDR was a good idea. Or, even worse, who the hell at Apple thought implementing that should happen by way of locking UI to the standard range. Even if you have a properly mastered HDR video (or image), and you've got your brightness set to where it doesn't hurt to look at, it still makes all the UI surrounding that image look grey. If I'm only supposed to watch HDR in fullscreen, where there's no surrounding UI, then maybe you should tone-map to SDR until I fullscreen the damn video?
crazygringo 46 minutes ago [-]
Yup, totally agreed. I said the same thing in another comment -- HDR should be reserved only for full-screen stuff where you want to be immersed in it, like movies and TV shows.

Unless you're using a video editor or something, everything should just be SDR when it's within a user interface.

hbn 2 hours ago [-]
HDR videos on social media look terrible because the UI isn’t in HDR while the video isn’t. So you have this insanely bright video that more or less ignores your brightness settings, and then dim icons on top of it that almost look incomplete or fuzzy cause of their surroundings. It looks bizarre and terrible.
crazygringo 38 minutes ago [-]
The alternative is even worse, where the whole UI is blinding you. Plus, that level of brightness isn't meant to be sustained.

The solution is for social media to be SDR, not for the UI to be HDR.

NathanielK 2 hours ago [-]
It's good if you have black text on white background, since your app can have good contrast without searing your eyes. People started switching to dark themes to avoid having their eyeballs seared monitors with the brightness high.

For things filmed with HDR in mind it's a benefit. Bummer things always get taken to the extreme.

hombre_fatal 53 minutes ago [-]
Not sure how it works on Android, but it's such amateur UX on Apple's part.

99.9% of people expect HDR content to get capped / tone-mapped to their display's brightness setting.

That way, HDR content is just magically better. I think this is already how HDR works on non-HDR displays?

For the 0.01% of people who want something different, it should be a toggle.

Unfortunately I think this is either (A) amateur enshittification like with their keyboards 10 years ago, or (B) Apple specifically likes how it works since it forces you to see their "XDR tech" even though it's a horrible experience day to day.

nine_k 2 hours ago [-]
But isn't it the point? Try looking at a light bulb; everything around it is so much less bright.

OTOH pointing a flaslight at your face is at least impolite. I would put a dark filter on top of HDR vdeos until a video is clicked for watching.

bofaGuy 2 hours ago [-]
Netflix has been the worst performing and lowest quality video stream of any of the streaming services. Fuzzy video, lots of visual noise and artifacts. Just plan bad and this is on the 4k plan on 1GB fiber on a 4k Apple TV. I can literally tell when someone is watching Netflix without knowing because it looks like shit.
mapontosevenths 1 hours ago [-]
It's not AV1's fault though, I'm pretty sure it's that they cheap out on the bitrate. Apple is among the highest bitrates (other than Sony's weird hardware locked streaming service).

I actually blamed AV1 for the macro-blocking and generally awful experience of watching horror films on Netflix for a long time. Then I realized other sources using AV1 were better.

If you press ctl-alt-shift-d while the video is playing you'll note that most of the time that the bitrate is appallingly low, and also that Netflix plays their own original content using higher bitrate HEVC rather than AV1.

That's because they actually want it to look good. For partner content they often default back to lower bitrate AV1, because they just don't care.

odo1242 1 hours ago [-]
This is actually their DRM speaking. If you watch it on a Linux device or basically anything that isn’t a smart TV on the latest OS, they limit you to a 720p low bitrate stream, even if you pay for 4k. (See Louis Rossman’s video on the topic)
jsheard 1 hours ago [-]
OP said they're using an Apple TV, which most definitely supports the 4K DRM.
array_key_first 25 minutes ago [-]
The bit rate is unfortunately crushed to hell and back, leading to blockiness on 4K.
pcchristie 8 minutes ago [-]
I cancelled Netflix for this exact reason. 4K Netflix looks worse than 720 YouTube, yet I pay(paid) for Netflix 4K, and at roughly 2x what I paid for Netflix when it launched. It's genuinely a disgrace how they can even claim with a straight face that you're actually watching 4K. The last price rise was the tipping point and I tapped out after 11 years.
mulderc 41 minutes ago [-]
I also find Netflix video quality shockingly bad and oddly inconsistent. I think they just don’t prioritize video quality in the same way as say apple or Disney does.
mtoner23 1 hours ago [-]
Probably some function of your location to data centers. I find hbo max to be aysmal these days. But I've learned to just stop caring about this stuff since no one else in my life does
jiggawatts 18 minutes ago [-]
https://xkcd.com/1015/

Now you can be mad about two things nobody else notices.

Eduard 2 hours ago [-]
I'm surprised AV1 usage is only at 30%. Is AV1 so demanding that Netflix clients without AV1 hardware acceleration capabilities would be overwhelmed by it?
FrostKiwi 2 hours ago [-]
Thanks to libdav1d's [1] lovingly hand crafted SIMD ASM instructions it's actually possible to reasonably playback AV1 without hardware acceleration, but basically yes: From Snapdragon 8 onwards, Google Tensor G3 onwards, NVIDIA RTX 3000 series onwards. All relatively new .

[1] https://code.videolan.org/videolan/dav1d

jeffparsons 25 minutes ago [-]
It's possible without specific hardware acceleration, but murderous for mobile devices.
snvzz 2 hours ago [-]
Even RISC-V vector assembly[0].

0. https://code.videolan.org/videolan/dav1d/-/issues/435

johncolanduoni 13 minutes ago [-]
Compression gains will mostly be for the benefit of the streaming platform’s bills/infra unless you’re trying to stream 4K 60fps on hotel wifi (or if you can’t decode last-gen codecs on hardware either ). Apparently streaming platforms still favor user experience enough to not heat their rooms for no observable improvement. Also a TV CPU can barely decode a PNG still in software - video decoding of any kind is simply impossible.
adgjlsfhk1 2 hours ago [-]
There are a lot of 10 year old TVs/fire sticks still in use that have a CPU that maxes out running the UI and rely exclusively on hardware decoding for all codecs (e.g. they couldn't hardware decode h264 either). Image a super budget phone from ~2012 and you'll have some idea the hardware capability we're dealing with.
eru 2 hours ago [-]
If you are on a mobile device, decoding without hardware assistance might not overwhelm the processors directly, but it might drain your battery unnecessarily fast?
boterock 2 hours ago [-]
tv manufacturers don't want high end chips for their tv sets... hardware decoding is just a way to make cheaper chips for tvs.
resolutefunctor 2 hours ago [-]
This is really cool. Props to the team that created AV1. Very impressive
conartist6 37 minutes ago [-]
For a second there I wasn't looking very close and I thought it said that 30% of Netflix was running on .AVI files
tr45872267 2 hours ago [-]
>AV1 sessions use one-third less bandwidth than both AVC and HEVC

Sounds like they set HEVC to higher quality then? Otherwise how could it be the same as AVC?

pornel 2 hours ago [-]
There are other possible explanations, e.g. AVC and HEVC are set to the same bitrate, so AVC streams lose quality, while AV1 targets HEVC's quality. Or they compare AV1 traffic to the sum of all mixed H.26x traffic. Or the rates vary in more complex ways and that's an (over)simplified summary for the purpose of the post.

Netflix developed VMAF, so they're definitely aware of the complexity of matching quality across codecs and bitrates.

tr45872267 2 hours ago [-]
I have no doubt they know what they are doing. But it's a srange metric no matter how you slice it. Why compare AV1's bandwith to the average of h.264 and h.265, and without any more details about resolution or compression ratio? Reading between the lines, it sounds like they use AV1 for low bandwidth and h.265 for high bandwidth and h.264 as a fallback. If that is the case, why bring up this strange average bandwidth comparison?
dylan604 2 hours ago [-]
definitely reads like "you're holding it wrong" to me as well
ls612 2 hours ago [-]
On a related note, why are release groups not putting out AV1 WEB-DLs? Most 4K stuff is h265 now but if AV1 is supplied without re-encoding surely that would be better?
avidiax 2 hours ago [-]
I looked into this before, and the short answer is that release groups would be allowed to release in AV1, but the market seems to prefer H264 and H265 because of compatibility and release speed. Encoding AV1 to an archival quality takes too long, reduces playback compatibility, and doesn't save that much space.

There also are no scene rules for AV1, only for H265 [1]

[1] https://scenerules.org/html/2020_X265.html

breve 19 minutes ago [-]
> Encoding AV1 to an archival quality takes too long

With the SVT-AV1 encoder you can achieve better quality in less time versus the x265 encoder. You just have to use the right presets. See the encoding results section:

https://www.spiedigitallibrary.org/conference-proceedings-of...

ls612 2 hours ago [-]
Yeah I’m talking about web-dl though not a rip so there is no encoding necessary.
chrisfosterelli 1 hours ago [-]
Player compatibility. Netflix can use AV1 and send it to the devices that support it while sending H265 to those that don't. A release group puts out AV1 and a good chunk of users start avoiding their releases because they can't figure out why it doesn't play (or plays poorly).
mrbluecoat 9 minutes ago [-]
h.264 has near-universal device support and almost no playback issues at the expensive of slightly larger file sizes. h.265 and av1 give you 10-bit 4K but playback on even modest laptops can become choppy or produce render artifacts. I tried all three, desperately wanting av1 to win but Jellyfin on a small streaming server just couldn't keep up.
Dwedit 2 hours ago [-]
Because pirates are unaffected by the patent situation with H.265.
ls612 2 hours ago [-]
But isn’t AV1 just better than h.265 now regardless of the patents? The only downside is limited compatibility.
BlaDeKke 1 hours ago [-]
Encoding my 40TB library to AV1 with software encoding without losing quality would take more then a year of not multiple years, consume lots of power while doing this, to save a little bit of storage. Granted, after a year of non stop encoding I would save a few TB of space. But it think it is cheaper to buy a new 20TB hard drive than the electricity used for the encoding.
phantasmish 39 minutes ago [-]
I avoid av1 downloads when possible because I don’t want to have to figure out how to disable film grain synthesis and then deal with whatever damage that causes to apparent quality on a video that was encoded with it in mind. Like I just don’t want any encoding that supports that, if I can stay away from it.
Wowfunhappy 1 minutes ago [-]
With HEVC you just don't have the option to disable film grain because it's burned into the video stream.
adgjlsfhk1 5 minutes ago [-]
What's wrong with film grain synthesis? Most film grain in modern films is "fake" anyway (The modern VFX pipeline first removes grain, then adds effects, and lastly re-adds fake grain), so instead of forcing the codec to try to compress lots of noise (and end up blurring lots of it away), we can just have the codec encode the noisless version and put the noise on after.
coppsilgold 7 minutes ago [-]
In MPV it's just "vf=format:film-grain=no" in the config. And I prefer AV1 because of this, everything looks better without that noise.
shmerl 36 minutes ago [-]
Qualcomm seems to be lagging behind and doesn't have AV1 decoder except in high end SoCs.
badmonster 1 hours ago [-]
[dead]
beritdotdev 1 hours ago [-]
[dead]
kvirani 2 hours ago [-]
Top post without a single comment and only 29 points. Clearly my mental model of how posts bubble to the top is broken.
yjftsjthsd-h 2 hours ago [-]
IIRC, there's a time/recency factor. If we assume that most people don't browse /newest (without commenting on should, I suspect this is true), then that seems like a reasonable way to help surface things; enough upvotes to indicate interest means a story gets a chance at the front page.