Hacker News Reader

Youtube already has it's own HDR video codec with VP9. Also, HDR10+ and Dolby Vision aren't HDR formats, they're formats for storing dynamic metadata that can help TVs better display HDR video. The article seems to misinterpret what their purpose is. HDR video can be presented just fine without HDR10+ or Dolby Vision.

Thanks!

> HDR video can be presented just fine without HDR10+ or Dolby Vision.

To see HDR content at its full dynamic range, you'll need an HDR-capable device or display. Viewers watching on non-HDR devices/displays will see an SDR¹ video derived from the HDR source.

¹ "Standard dynamic range" or "smooshed dynamic range"

> To see HDR content at its full dynamic range, you'll need an HDR-capable device or display.

Not exactly - you need an HDR Mastering display to see HDR content at full dynamic range. There are essentially no high volume consumer-level devices, with the exception of maybe Apple's XDR lineup (MBP, iPad, iPhone, Pro Display) with the capability of displaying non-windowed HDR content at full brightness.

Everything else relies on tone mapping, even the latest 2022 OLED & QDOLED TVs.

Even HDR mastering monitors can only reach 2000 to 4000 nits [1], whereas PQ gamma goes up to 10000 nits. This is why most hdr streams contain metadata containing mastering display peak luminance.

[1] https://www.usa.canon.com/shop/p/dp-v3120

Good info, thanks! Can you elaborate on why you qualified this with "non-windowed"?

So if VP9 and AV1 already store HDR data, How is Dolby Vision used? Its HDR metadata shipped alongside video formats that don't already encode this data?

Like if I stream netflix to a TV supporting Dolby Vision what format is the video being streamed in and is the TV manufacturer just paying Dolby for the right to correctly decode this HDR info then?

Dolby Vision is a format of that metadata - it's hiding either at the beginning or as part of a frames in proprietary metadata extensions. When the video is being played, this metadata is extracted by video decoder (or demuxer) and then sent together with video frames to the display where the display then applies the (color, etc.) metadata to correctly show the frames based on its capabilities.

Since the format of this metadata is proprietary, the demuxers, decoders and displays need to understand it and properly apply it when rendering. That's the part that needs to be implemented and paid for.

But that's really not all of the technology - Dolby Vision isn't just the metadata format, it's also definition of how the videos are mastered and under which limitations (e.g. DV allows video to be stored in 12-bit per pixel format, allows mastering with up to 10.000 nits of brightness for white pixels and defines wider color range so better, brighter colors can be displayed by a panel capable of doing that).

https://www.elecard.com/page/article_hdr is actually a pretty good article that overviews this topic (although you do need a basic understading how digital video encoding works).

That we restrict who can properly decode data is even more absurd than other software patents.

Dolby Vision supplements the HDR data already present in a video. For example, when you buy a Blu-ray disc that supports Dolby Vision, the disc contains several m2ts files containing HEVC encoded videos. Present within the HEVC stream is also the metadata, which supplies metadata for each frame. To see what this data is, I used[1], and then got the info for frame 1000 and saved it as a json in this [2] pastebin. As you can see, it contains info regarding the minimum and maximum PQ encoded values, the coefficients of the ycc to rgb matrix, among other things. This allows TVs to better display the HDR data, as currently, video data encoded using rec. 2020 color primaries with max light level of 10000 nits is far outside what current TVs are capable of displaying, so metadata showing max pq of a frame or scene allows these devices to make better decisions.

[1] https://forum.doom9.org/showthread.php?t=183479 [2] https://pastebin.com/m5NfUTbc

Linus Tech tips talked about it. HDR on YouTube is a nightmare. There's no way to know if it'll come out in the right colours or get crushed into an ugly mess.

That’s kind of false.

The entire points of Dolby Vision is that it should give you the same image no matter what display you display it on, if the device is capable of displaying it. It’s an absolute format of light and color values, rather than a format based on RGB values.

> I’ll welcome anyone that wants to enter the space that thinks they can do better,

"Do better" is tricky to define. By some metrics, Ambisonics, a decades-old, license free technology, "does better" than Atmos does. But by others, it does worse. Which metrics are important?

Content is King here. Do you have source material and do you have playback devices or software out in the world?

Google has both with YouTube, Chromecasts and Android phones and TVs. They’re one of the few power players that can unilaterally change up their codec and metadata suite, but only as far as YouTube goes.

So “do better” means getting enough content behind a tech stack and still delivering a satisfying experience to the customer. If they can meet or exceed what Dolby delivers, I think that would be great! Even if they only match Dolby, that’s still pretty good.

That's a good point.

If support for playing Ambisonics was added to (say) vlc, and easy-to-use conversion tools from Atmos to Ambisonics also existed, that'd probably go a long way to increasing it's adoption.

Ambisonics hardware was on sale in the late 1970s and early 80s, but few people at that time had multi-speaker setups, which meant that the companies making it went defunct.

Libre and gratis ambisonics encoding and decoding software has existed for more than a decade. It is the domain of enthusiasts rather than regular consumers.

Makes sense, and this is actually the first I've heard of ambisonics. From your description, it doesn't sound like the kind of thing there will ever be tons of content for, so not tons of demand either. But a small demo-scene-like thing can be neat too.

I think the most important figure of merit is sadly how much money it will make the people in charge of implementing it in consumer devices.

> I suspect what will happen is YouTube will develop their own HDR and audio codecs

It is stated in the article they're backing HDR10+.

Not sure about the audio.

Yeah there’s some supposition in the article so I’m not clear if it’s HDR10+ or an extension of it, just that Google wants to do something to break Dolby’s bank and they’ve got a heist crew with something to prove (i.e. Netflix, Samsung, Meta, et al.) and a placeholder name for the audio stuff (“Immersive Audio Container”). The distinction hardly matters since it’ll just be one more format (or set of formats) for manufacturers to support.

For audio they are working on something new called Immersive Audio Container

https://aomediacodec.github.io/iac/

This could put pressure on Dolby to release their older stuff for free (no licensing) which would be a huge win for everyone. But I agree about Google - the whole point of Dolby is to have a high quality standard that is the industry choice for consistency. That does cost money (mainly in licensing their stuff or chips that use their encoder) but the way I see they have to make profit somehow. Is it overpriced? Probably.

Is older standard dolby digital (and dolby digital plus) 5.1 surround sound still pretty damn good? Yep - and it should be free. They have 20 years of newer, superior stuff to make money from!

AC-3’s (Dolby Digital) patents all expired in 2017. I’m not sure about E-AC-3 (Dolby Digital Plus), but my understanding is producers just move onto new sound technology. It’s not that DD and DD+ don’t both sound great, it’s that people move with advances in sound production to stay on or near the state of the art. If you want to write an AC-3 encoder/decoder, go for it, but that’s not much help for folks that want to use AC-4, TrueHD and Atmos.

Well the reason I mention those two older formats is that those are the encoders you need to use a TON of older but still very useable hifi equipment. As far as I know if you want to get good quality sound to anything that uses TOSLINK (very common connector not just on older equipment) or the single orange digital audio cable (that looks like an orange RCA port) too. I specifically bought an older Roku 4K model that has the dolby digital plus encoder chip built in (and a TOSLINK output) so that I can keep using my still very nice sounding Yamaha receiver from the early 2000's. They only had the TOSLINK and DD chip for a brief period in the Roku 4K (current one does not have it and the oldest ones do not either). Basically what this means is that Netflix can have the latest and greatest audio on their stream (whatever is the best of the best in 5.1 currently) and this little chip will nicely re-encode it on the fly down to the best quality format my receiver can handle. If you do HDMI passthrough to Toslink or similar it will not work as theres no chip to re-encode.

Even a ton of modern motherboards have TOSLINK and lots of new equipment as well so its a worthwhile way to get a baseline quality 5.1 audio that still holds up very, very well for home systems today. DD+ with a good receiver and large speakers will blow away most of the cheaper Atmos systems.

You could also theoretically do this with most computers or a laptop with the right hardware and a little software digital encoder but the issue is that most of the time the way they have instituted DRM means that the browser is not going to even have access to 5.1 in the first place or your device (like any Nvidia graphics card for example - even if the HDMI output can include high quality audio) is going to just get 2.1 audio.

Speaking of Stereo Audio vs Encoding Schemes, there was a great article posted here a while back about diving into audio formats on 35mm movie film [0]. There's a photo showing the analog stereo wave-form alongside two digital tracks: Dolby and Sony [1]. So the audio is physically printed onto the film in 3 different formats (1 analog, 2 digital) and it's up to the projector to decide which one is needs/has-the-hardware-for.

[0] https://news.ycombinator.com/item?id=30919904

To complement what you say, you also have to consider that there are an absolute myriad of processes, people and companies within the audiovisual industry where the Dolby workflow is almost inseparable from what they do and what they work with. People and companies that have made investments, sometimes of many zeros and many hours, to be able to work this way.

Dolby is more than a standard or spitting six channels of raw audio out of speakers. That's just the end product, because it's not just standard itself.

It is, for example, the hardware that is in cinemas and home theaters (encoders, decoders, Dolby RMU...), the certification processes that Dolby does in cinemas and recording and mixing studios, the mixing technicians who work with all that and send the final mixes with the netflix/hbo/whatever specifications, vendors, integration partners, speaker manufacturers...

There are also plugins that work in DAWs like Pro Tools, the ecosystem (Dolby Atmos Renderer, Dolby Atmos Production Suite), just to scratch the surface.

One thing is to publish a standard and another the ecosystem around that standard. It is interesting that there are new standards, but given Google's history with its long term attention span I have my doubts that this will materialize into anything more than an internal asset for google.

> and ultimately this will be to increase the quality of YouTube while minimizing their licensing costs.

Yes, I can also imagine they have specific requirements on the file format like quickly skipping to timestamps, highly variable bitrate, handling text and graphics well etc. I imagine their requirements to be so general that it'll benefit anyone, especially those that do streaming.

In either case "just one more standard" (or relevant xkcd) is an unavoidable obstacle for every new standard, and does not mean the project will fail. I have lots of critique against Google but this is one thing they are positioned to do well, and have a decent track record. And given how the competition operates, is frankly refreshing.

Agreed.