W3C

– DRAFT –
Media and Entertainment IG

05 November 2024

Attendees

Present
Andreas_Tai, Bernd_Czelhan, Chris_Needham, Francois_Daoust, Hiroki_Endo, Hisayuki_Ohmata, Ingar_Arntzen, Kaz_Ashimura, Louay_Bassbouss, Niko_Faerber, Piers_O'Hanlon, Ryo_Yasukawa, Tatsuya_Igarashi
Regrets
-
Chair
Chris, Igarashi
Scribe
cpn, kaz, tidoust

Meeting minutes

Introduction

Chris: Welcome, in this meeting we'll recap some of the main outcomes from TPAC and talk about next steps.
… Also a discussion around the Multi-Device Timing CG, where I saw it was proposed to be closed.
… Anything else?

(nothing)

Slideset: https://www.w3.org/2011/webtv/wiki/images/c/cb/MEIG-2024-11-05.pdf

Next Generation Audio and Video Codecs

Chris: NGA proposal presented at TPAC MEIG meeting.
… I want to figure out where to go next with it.
… Outcomes from the meeting: gap analysis for audio, also consider video.
… Also there was opportunity at TPAC to talk with people outside of the meeting.

Bernd: Not quite sure about the gap analysis.

Chris: Let's figure out what it really means
… We can start with use cases.
… Some of this in the document we prepared and you presented at TPAC, so we have some material already.

Bernd: Should we come with a proposal? Learn from the industry what's missing on the web compared to other platforms. We're in contact.
… What's done so far and what's still missing to deliver something which is a good user experience.
… Have some service provider to show there's the need.

Chris: This is a conversation TV and media industry is having elsewhere?

Bernd: My understanding is maybe some service provider can address today's technology
… but I'm missing how should we work on that.

Chris: We could write a document to capture the points
… that kind of use cases and gap analysis
… that's what we can turn into a document
… more technical analysis too

Bernd: And have a meeting for that purpose?

Chris: Yeah, we can do that.
… We can use this meeting also, or can have special topic-focused meetings.
… If we have a lot to talk about in the main MEIG meeting, we can schedule additional meetings.
… We can be flexible.

Bernd: Sounds perfect.
… We can already work on the document,
… also dedicated meeting would be good.

Chris: Sounds good to me.
… Happy to talk with you separately about getting the media industry input needed.

Bernd: ok.

Chris: I'm not sure about next-generation video at the moment, I don't know the status of development.
… That would be interesting, to hear from somebody with expertise to give an introduction to the technology
… could also cover that
… if you're designing APIs
… bigger problem with focusing on
… but I think moving up would really require special knowledge
… might be longer term
… was interesting a couple of people showed interest
… think some of the materials are already there

Bernd: yeah
… some requirements there
… real deployment would be helpful

Chris: yeah, industry support would be good
… any other thoughts?

Niko: One question, a gap analysis describes a difference between two points.
… What are those two points?

Chris: I see two ways to approach this. One is between the user needs, and the current technical capability the Web has,
… so starting from the use case perspective, looking at what's supported by browsers today and documenting what's missing.
… The alternative, and the approach we've taken so far, is more codec-centric.
… We have some particular technology that enables use cases, so we look at how to bring that to the web.
… So I think we need some use cases and do the former analysis, e.g., what can be done with Web Audio today.
… We have some of this already.

Niko: ok, that matches my understanding from TPAC.

Chris: We can follow up this separately on the industry side.

Media WG update

Chris: A few updates.
… Media WG just agreed to publish the Audio Session API spec.
… It allows websites to control how multiple audio sources get mixed, e.g., music and notifications, etc.
… The music volume might be temporary adjusted so you hear the notification. Feedback on the spec is welcome.
… WebCodecs, VideoFrame metadata for camera effects, e.g., background blur
… A lot of the development is driven by RTC use cases, very low latency streaming.
… Camera effects is an example. Do we need to talk about requirements from media creation use cases?
… Finally, there was some agreement at TPAC to look at Text Tracks in MSE,
… so support for timed text carried in media container files or the video bitstream.
… I'm hoping to organize a dedicated session on that soon.
… It's something my organization is interested in.
… Any thoughts/questions?

(nothing)

Multi-Device Timing

Chris: This has been a long term topic, discussion at TPAC 2016.

Ingar: did some work to generate interesting forum around timing
… starting at TPAC 2016

https://www.w3.org/2016/09/22-sync-minutes.html minutes

Ingar: We put this into a standardization draft, just a pending some interest.
… Should I say something about the initiative itself?

Chris: Yes

Ingar: Opportunity with the web, to have timing. It's a collection of dynamic documents.
… Think about it as a thing for rendering data, visualisations.
… Anything could be sync'd up and played back together. Multimedia frameworks would support it, but the web doesn't have that.
… Hard to use them together. Idea of timing object is to make a generic spec for a timing model, timing control and timeline,
… and make it independent of specific frameworks.
… It's been successful, we use it a lot, to solve many problems.
… There's interest in syncing things to video or audio.
… The missing piece isn't much. For what should be standardised: not a lot.
… The thing that would matter is the media element having a mode for sync. Media element wasn't made for that,
… e.g., if you ask it to skip it'll spend some time doing that, then you'd need to sync again.
… Solution to have a sync mode in the media element. Instead of passing a command, pass a timestamp vector to aim for.
… It would open up the important use cases.
… The multi-device aspect isn't necessary. Timing is more important.
… Not only entertainment media, also recording and replaying sensor data.
… Shame there was no interest from browsers.

Chris: At TPAC 2023 Apple presented the Media Session Coordinator proposal. The use case is co-viewing of media content.
… Not at the same timing accuracy level as Timing Object enables.
… But even that didn't get traction with other Web browser vendors.

Ingar: That's just one use case for timing,
… but three are thousands of use cases there.
… Media Session Coordinator bundles together timing with session management,
… whereas Timing Object focsues on just the sync part

Chris: Yes, those things are kind of orthogonal.

Ingar: Right

Kaz: I've been working with MEIG for a while, and also WoT group, and the multi-modal group 10 years ago.
… Do we need to think about synchronisation just for web browsers, or maybe also non-browser entities like IoT sensors, surveillance cameras?
… Could use exsting statechart mechanisms also?

Ingar: It's not specific to the web, yes. Even though we named it "multi-device timing", we didnt' really look at the multi-device part, more the timing object.
… Having a standard way to do the multi-device part could be a consideration. But the timing part opens up the use cases.

Kaz: During TPAC2024, Audio WG people were intersted in precise sync. Could create a document with use cases.

Ingar: Sure, the findings back then, the limiting factor was the ability to locally control the media element. We got that down to around 7-10ms on reasonable devices
… For general sync of timing objects across web clients, was 1 or 2 ms. Good enough for a lot of applications, or maybe not for some specific audio apps.

Kaz: Technically, need some precise base clock.

Ingar: You need a clock, and can rely on the server you connect to. That's the way we did it, and sync based on network latency, you can get to 1-2 ms off.
… You shouldn't have to depend on anything, the important thing is availabiltiy of a sync signal, and see the clock as an app level object and not an infrstructure resource

Kaz: So Timing Object is like a virtual clock?

Ingar: It's like a logical clock, stopwatch.

TPAC Breakout: Sync on the Web

Chris: There was related breakout on sync on the Web
… Someone mentioned the Web Audio conference, which sometimes discusses millisecond-accurate synchronization.
… APIs for timestamps or latency for input-generating APIs, Gamepad, USB, MIDI, to enable the app to do synchronization.
… Some points made for digital twins as use case, cloud gaming as well
… Discussion on that happens within the Web & Networks IG

Ingar: There's overlap there, as long as you have a timing object for event handling
… sometimes people monitor the latency, can be very complicated but could simplify it.

Chris: A broad question to the group. Is there a renewed interest in this topic, and something to follow up?
… Ingar, you have solutions that are working pretty well.
… Maybe there's not so much to be done here?
… It seems a shame to close the CG...

Ingar: got a report to sending the content accessible

Chris: All the archives can still be accessed if the group closes.

Ingar: There is still an opportunity there,
… one thing you could do is going back to the industry to find people interested
… many type of application domains.

Chris: We heard from Mr. Komatsu from NTT, on how to do sync using Media over QUIC.
… This "Sync on the Web" session was a reminder to think about the topic more generally.
… We should let him know about this discussion and see what he wants to do next.

Ingar: We talk to industry from time to time, a media use case is the lottery show, with video and synchronised text and graphics.
… It seems such an easy thing, but that's pretty hard for developers to do today, and special cases with media players on different platforms.
… It's really a showstopper for media industry business.

Francois: Ian and Dom are handling the CG procedure and checking the progress of CGs.
… I wanted to confirm that closing the CG itself would not lose the content, nothing would be lost.
… We could write a blog post to summarise and point people to MEIG for future discussion when we revive the topic.
… I also wanted to note we tried for years to push browsers to consider the spec and the approach.
… The feedback has always been to expose hooks, latency measures, but the leave orchestration up to the application.

Chris: A potential improvement could be around the currentTime and playbackRate attributes?
… The HTML spec doesn't set any accuracy or requirements.
… And the ability for browsers might vary by platform, e.g., desktop, mobile or TV.
… This may not be a spec issue, more about quality of implementation.

Ingar: even the sort of properties
… is that good enough?
… skipping is happening
… that is always a source of noise

Chris: Focusing on that level might be productive. Show the practical data we have on accuracy, raise browser issues.

Ingar: Just having a sort of test with some kind of video content might be useful.

Louay: Regarding testing, I'm working on the CTA WAVE test suite.
… Video playback tests include the current time.
… there is a feature on Vision Pro devices called SharePlay.
… You can select any applications to listen to music or watch video and synchronize.

Chris: Yes, Media Session Coordinator was essentially bringing that feature to the web.

Ingar: It's important to open up it for anybody, so it's not limited to any vendor specific protocols.

kaz: Another discussion place is Web of Things. I'll talk with WoT chairs also. Time synchronization is not yet defined within the model.

Content Authenticity

Chris: Another topic from TPAC was Content Authenticity: Originator Profile and C2PA, which we've discussed here before.
… Discussions are happening about a potential workshop next year.

CTA WAVE Collaboration

Chris: We had a meeting in September and at TPAC about their streaming media test suite.
… We're planning to have a follow up meeting to look at some MSE specific issues, hopefully soon.

Next meeting

Chris: Same time next month: 3 Dec at 15:00 UTC.

[adjourned]

Minutes manually created (not a transcript), formatted by scribe.perl version 196 (Thu Oct 27 17:06:44 2022 UTC).