W3C

Timed Text Working Group Teleconference

04 June 2026

Attendees

Present
Atsushi, Chris, Nigel, Pierre
Regrets
Andreas, Cyril, Gary
Chair
Nigel
Scribe
nigel, cpn

Meeting minutes

This Meeting

Nigel: [iterates through the agenda]
… One pull request on DAPT,
… IMSC 1.3 post-tidy-up tasks
… Media Type Definition and Profile Registry, AI, and TPAC planning

Anything else to add?

(Nothing)

IMSC 1.3 post-publication tidy-up tasks

Errata and Release Tag

Nigel: Where are we up to with the errata document and release tag?

Pierre: I created the release tag

Release: IMSC Text Profile 1.3 W3C Recommendation 21 May 2026

Nigel: Is there any action remaining on the errata?

Pierre: I don't know. There's a link to the errata, and it says there are none

Nigel: That's accurate

Atsushi: There's a need to update the errata script, it's just a copy of IMSC 1.2
… There should be a minor cosmetic update to the script, but it's there. I don't believe there's a default scheme covering entire W3C that's under maintanence, so not sure what will happen

Nigel: So I don't think there's any action for us to take

Atsushi: Unless about the cosmetic things

Nigel: The Rec says that errata exists. It's not great, but I'm not too upset

Atsushi: That's the default display in ReSpec

Nigel: I don't want to spend much time worrying about this

Atsushi: Maybe "Errata page exists" is more suitable?

Nigel: It's inconsistent with the rest of the header section, where everything is a URL and the label is the URL itself, whereas errata is different

Atsushi: That's a good issue to file against ReSpec

Nigel: I can do that

Pierre: Can we approve PR 645 on IMSC?
… It's about updating the readme

Nigel: Done

Pierre: There's another PR to return to ED after Rec publication. I believe we're ready to do that now
… Atsushi, it's ready to merge

Atsushi: There's a consideration about the public URL using a W3C name, for github.io publications

Nigel: That's fine
… The only thing that's awkward about it is that we really need a new shortname in the ED, as 1.3 is done. But until there's work to do we don't need a new shortname

Pierre: In the past, until we start new work, the head is the latest release. That's what I'd prefer to do

Nigel: In that case the github.io looks like the Rec and it's not the right URL for the Rec

Pierre: Is that bad? The converse seems worse, for there to be a new ED with changes when there aren't any

Atsushi: Even if content is identical as published as Rec, but published in github.io it should be an ED

Chris: The only risk seems to be that someone takes the github.io URL thinking it's
… the reference to the Rec

Pierre: The PR is merged now

Nigel: So it's done

IMSC 1.3 Namespace document

Chris: The namespace is done, so I've closed the issue.

TTML Profile Registry entry

Nigel: We agreed that we'd set up auto-publication of the Media Type Definition and Profile Registry and Note. Has that been set up?

Atsushi: I need to spend some time on it

Nigel: Pierre has approved it, so I think I won't wait for that to be done before merging the PR
… But it would be good to publish a new version of the Note as soon as possible
… The PR is merged
… I think all the IMSC actions are done. Is there anything else we need to do?

(Nothing)

Nigel: So we can drop IMSC from the agenda for future meetings unless something comes up

TTML Media Type Definition and Profile Registry

Nigel: There's still work on auto-publication to do
… For PRs 88 and 89, review would be helpful

Update references from TTML1 2nd Ed to 3rd Ed w3c/tt-profile-registry#88

github: w3c/tt-profile-registry#88

SUMMARY: Please review

Nigel: This updates the TTML1 2nd Ed reference to TTML1 3rd Ed, but I can see arguments why not to do that

Update DECE URLs to ones that work w3c/tt-profile-registry#89

github: w3c/tt-profile-registry#89

Nigel: Pierre, might be good for you to review. It's fixing the URLs for CFF that have been moved

Pierre: Does it match what we have in IMSC 1.3?
… The URLs work. Approved
… And they match IMSC 1.3

SUMMARY: @nigelmegitt to merge this PR

Impact of AI technologies on Timed Text Working Group's mission

Nigel: This came in the MEIG meeting on Tuesday (minutes: https://www.w3.org/2026/06/02-me-minutes.html)
… We discussed last meeting that some organisations use AI to generate additional data, e.g., emotion, or per-word timings, using AI
… and using those in their players. Someone mentioned wanting to do this in WebVTT, but you can't because it doesn't allow cue metadata
… They'd invented their own metadata cue format that can only be used in their own player
… I showed them an example using a namespace in TTML, where a player that understands the extensions would use them, and a player that doesn't would ignore them
… AI generated additional data that needs to be captured and maybe standardised in subtitle formats. The task is to understand what the requirements might be

Pierre: It might be simpler than that. The standards exist, and people don't know about them

Nigel: Useful to know if things are already possible

Pierre: Can do it in TTML with animations

Nigel: So could make a use case document to show it's possible in TTML

Pierre: Find out what library these people use, and make sure IMSC.js or whatever we use can support the additional styling, and I think we're done

Nigel: I reserve judgement. Examples I've seen it's not obvious you can do it, e.g, effects on a word by word basis

Pierre: I agree. Ultimately we might decide new TTML features are needed, I agree. But most of what I've seen is simple
… I'd start with solving their immediate problem. Lots of people are embracing AI but don't know what's possible with timed text and captions

Nigel: Two possible starting points: AI, or expressive captioning. Those things converge

Pierre: I'd look at the tools, and enable them to output TTML or a modern format, and have a renderer to PNG, something ffmpeg can use as input, put people on the right track

Nigel: I think it's worth going through the process to ask people what they want, then look at how
… I think people have some ideas about what they want to express, and how to do it

Pierre: We've heard from those folks, but they're distinct from people putting subtitles onto Tiktok or Instagram, who make things up, as they have no other option
… I'd focus on the people doing that today

Nigel: I don't know how to find those people, and their output is just video pixels...
… I'd propose to write a proposal for something to be done under the MEIG umbrella, so we get input from non W3C members more easily
… I took an action to write a task force proposal
… Your input is useful

Chris: One note from me is that MEIG is a member group.
… I use my Chair's discretion to invite people from outside,
… but we're not by default open to any non-members to participate.
… I think we would have to do the outreach.
… Regardless of the structure, even if we were fully open, we would still have to
… identify the stakeholders and invite them to contribute. They aren't necessarily
… going to show up by default.

Nigel: Absolutely. I know at least 2 organisations involved.
… any other thoughts on that?

Pierre: I would say that for me, the most efficient use of our time would be to make
… sure there is a way to render complex TTML animations and styling, so that they
… can be used with tools like ffmpeg and then promote that format to the various
… subtitle AI tool vendors and ask them to output in that format so they can be
… rendered in that cool way.

Chris: Dumb question. I've seen examples like on TikTok where they're rendered
… into the image. If you had WebVTT or IMSC captions would they support that?

Pierre: It's all done by hand using Premier or that kind of tool.

Chris: So it's working round a limitation in the TikTok player, for example?

Pierre: Exactly. My assumption is that if tools existed that could take a subtitle file
… and render it with all the capabilities they're looking for, that's what they would use.

Chris: Right, and it's not accessible to people.

Pierre: Right. I don't think people perceive that as an issue today, they just see
… the subtitles and captions as one part of the editorial process, but as a result
… they end up with something that isn't accessible.
… If there were a way for ffmpeg to render style-rich subtitles and captions then
… we would probably not be having this conversation because people would
… author in that caption format, basically, be it TTML or WebVTT.

Atsushi: In south-east and east Asia most video content is open captioned, and
… the closed captions are generated by AI from the actor's speech in the video,
… but most videos don't have closed captioned content especially for social media
… sites. I understand that some use an internal format for caption files,
… but in the end the output is just a movie without any closed caption files as far as I observe.

Nigel: I agree, there's still a lot who use the lowest bar thing they can, e.g., SRT or 608 because they perceive it ticks a legal box for them, but it's useless

Atsushi: I think something like autocaption for language around the world, so it might be possible to include captions for short form view
… This kind of activity could be initiated from Tiktok or south or east Asian companies

TPAC Planning

Nigel: TPAC is at the end of October this year. We've been asked to indicate if we want meeting time.
… Usually I'd say yes, but what would we talk about?
… I think we will ask for time, but how much?
… Does anyone have constraints on whether they'll be there

Pierre: Unlikely I'll be there

Nigel: This came up in MEIG on Tuesday, Chris asked about joint meetings, e.g., MEIG+TTWG and MEIG+TTWG+APAWG
… Advanced features might be good to talk about. And I want to come back to TTML2 2nd Edition. With DAPT nearly complete, no new requirements for TTML2 2nd Edition, so get it to Rec. There may be discussion around that
… Also Dana suggested talking about interop work on WebVTT
… Atsushi, will you be going?

Atushi: Not sure yet. Depends on budget

Meeting close

Nigel: Thanks all

Nigel: Next meeting is 2026-06-18

agenda for next meeting

Nigel: [adjourns meeting]

Minutes manually created (not a transcript), formatted by scribe.perl version 248 (Mon Oct 27 20:04:16 2025 UTC).