search

w3c / imsc

TTML Profiles for Internet Media Subtitles and Captions (IMSC)
https://w3c.github.io/imsc/
Other
31 stars 17 forks source link

[WR/ARIB] Compatibility with ARIB-TTML / 5. Additional style control #550

Open himorin opened 4 years ago

himorin commented 4 years ago

Per: https://github.com/w3c/ttwg/issues/116 Comment 3 (https://github.com/w3c/imsc/issues/545), 5

ARIB-TTML has the attributes to enhance its presentation capability. The “arib-tt:border” attribute specifies the border appearance of the region defined by an element such as div, p, and span. Similarly, the “arib-tt:marquee” attribute for scrolling a text, the “arib-tt:ruby” for ruby, the “arib-tt:letter-spacing” for manipulation of letter spacing, and “arib-tt:text-shadow” for controlling text shadow are available.

nigelmegitt commented 4 years ago

This looks like it might benefit from some close analysis (that I don't have time to do right now!) of the differences between the ARIB-TTML styling capabilities, those in TTML2 and those in HTML+CSS.

css-meeting-bot commented 4 years ago

The Timed Text Working Group just discussed [WR/ARIB] Compatibility with ARIB-TTML / 5. Additional style control imsc#550, and agreed to the following:

The full IRC log of that discussion <nigel> Topic: [WR/ARIB] Compatibility with ARIB-TTML / 5. Additional style control imsc#550
<nigel> github: https://github.com/w3c/imsc/issues/550
<nigel> Nigel: I commented that I think some analysis could be helpful. Any other thoughts?
<nigel> Pierre: I have not had time to carefully study that one.
<nigel> .. On letter spacing, I have actually had the opportunity to spend a lot of time on it.
<nigel> .. My understanding is that the exact same issue is present in digital cinema.
<nigel> .. Letter spacing is really important in all languages. The practice, in DC, is not to handle
<nigel> .. letter spacing in the font file itself, but to handle it at the markup level. I do not know why.
<nigel> .. Japanese cinema subtitles also allow precise letter spacing. This was a feature requested
<nigel> .. for TTML2 and ultimately rejected.
<nigel> .. The author will adjust kerning while authoring in their tool, and this will be reflected in
<nigel> .. the markup. I have asked many times why this couldn't be done in the font, especially
<nigel> .. since in the case of cinema a dedicated font file is provided with Japanese subtitles.
<nigel> .. I have never received an answer.
<nigel> .. I could understand if you could not provide a font file, I would say okay, it has to be
<nigel> .. handled as part of the markup, but if a bespoke font file is provided why not use that.
<nigel> Nigel: Do you know if there is a use case for different kerning between the same characters
<nigel> .. in different parts of the same presentatoin?
<nigel> s/oin/ion
<nigel> Pierre: I did ask precisely that and did not get an answer.
<nigel> Gary: I wonder if this is partially something to do with direction of text, where particularly
<nigel> .. in Japanese you want to adjust it more because if you adjust it one way then the other
<nigel> .. way will be wrong, too large or too small. I'm not sure how often that is actually the case
<nigel> .. for captions but it could happen on the web.
<nigel> Nigel: Is there CSS for this?
<nigel> Gary: Yes, letter-spacing property
<nigel> .. It just takes a length
<nigel> Nigel: In the context of the web, specifying this is current practice.
<nigel> .. It's in CSS 1, SVG, CSS 2.1. There's also font-kerning, which sets the use of the kerning
<nigel> .. information held within the font.
<nigel> .. That's in CSS Fonts Level 3 CR
<nigel> .. I don't remember why we rejected this in TTML2.
<nigel> Pierre: [looks for it] It's #52.
<nigel> Nigel: Also #118
<nigel> Nigel: I see #52 came from the tracker, was raised there by Pierre and originated in a SMPTE liaison.
<nigel> .. And then #118 was also from the tracker, raised by Glenn, and originated in ARIB-TT.
<nigel> .. So we have seen this before, but there's no record of it.
<nigel> .. I see that we did add tts:letterSpacing! So this is resolved.
<nigel> Pierre: Please correct my earlier statement - letterSpacing is in TTML2.
<nigel> Nigel: Now what I want to know is what is different about ARIB-TT's letter spacing from what
<nigel> .. we have in TTML2?
<nigel> SUMMARY: More work needed to understand any semantic differences between similar features in ARIB-TT and TTML2
css-meeting-bot commented 4 years ago

The Timed Text Working Group just discussed [WR/ARIB] Compatibility with ARIB-TTML / 5. Additional style control imsc#550 [continued].

The full IRC log of that discussion <nigel> Topic: [WR/ARIB] Compatibility with ARIB-TTML / 5. Additional style control imsc#550 [continued]
<nigel> github: https://github.com/w3c/imsc/issues/550
<nigel> Pierre: I recall that letter spacing is not supposed to be used on a character by character basis.
<nigel> .. ipd is supposed to allow character by character adjustment, and it is specified not to be
<nigel> .. negative in TTML2. So I'm 99% certain that the ARIB-TT feature maps to ipd not letterSpacing.
<nigel> .. We need to study this in more detail but I wanted to add this for the record.
css-meeting-bot commented 4 years ago

The Timed Text Working Group just discussed [WR/ARIB] Compatibility with ARIB-TTML / 5. Additional style control imsc#550, and agreed to the following:

The full IRC log of that discussion <nigel> Topic: [WR/ARIB] Compatibility with ARIB-TTML / 5. Additional style control imsc#550
<nigel> github: https://github.com/w3c/imsc/issues/550
<nigel> Pierre: Can we come back to the character space question?
<nigel> .. If I'm not mistaken ARIB-TT letter spacing is to control the space between individual
<nigel> .. letters, positive and negative. The TTML2 feature tts:letterSpacing is intended to be applied
<nigel> .. to a sequence of characters to grow or shrink the space between characters overall, rather
<nigel> .. than between individual characters. For that there is tts:ipd, but that does not support
<nigel> .. negative values.
<nigel> .. We should confirm that ARIB's use of letter spacing is for individual character pairs.
<nigel> Nigel: Is there anything to stop you using TTML2 tts:letterSpacing and putting 2 characters
<nigel> .. in it though?
<nigel> Pierre: Imagine you have 3 characters and you want to change the spacing between each of
<nigel> .. the two pairs. You can't do that with tts:letterSpacing.
<nigel> Nigel: This is bringing back nasty memories
<nigel> Pierre: I'm fairly certain we concluded that we could not support controlling negative
<nigel> .. spacing between pairs of letters independently.
<nigel> .. The reason I remember this is that the request came from Digital Cinema originally
<nigel> .. where there is something like IPD that supports negative values. I think that may have
<nigel> .. been introduced specifically to support the Japanese market.
<nigel> Cyril: If you're nesting spans and changing the letterSpacing at each span level would it
<nigel> .. disallow you from controlling the spacing between each pair?
<nigel> Pierre: Right, but in a long sentence you end up with a terrible hierarchical structure.
<nigel> Cyril: I know!
<nigel> Nigel: I don't remember the answer but I remember asking if letterSpacing applies to the
<nigel> .. distance between the last character of the previous span and the first character of the
<nigel> .. span with letterSpacing set, or conversely at the end.
<nigel> Pierre: I'm still struggling to understand why the right solution is not to handle bespoke
<nigel> .. spacing in the font itself.
<nigel> Nigel: It's orthogonal though, you might want to change the letter spacing for one instance
<nigel> .. of a word only, e.g. the word WAVE in the TTML2 example. You can still do it on the font as well, for all uses.
<nigel> Pierre: That would be the perfect use cases, for artistic purposes, where there is an exceptional
<nigel> .. reason, negative IPD could be really helpful. But the reality might be that people would
<nigel> .. use it between every character, which is what I've seen in digital cinema, which defeats
<nigel> .. the purpose entirely, in my mind.
<nigel> Nigel: [wonders aloud about ipd or letterSpacing on empty spans]
<nigel> Pierre: ipd is really simple, that's done I think.
<nigel> Cyril: What happens if you set an ipd that is smaller than the calculated ipd? Does it limit
<nigel> .. the dimension?
<nigel> Pierre: Say that again?
<nigel> Cyril: Say you have a span, and you want to reduce the advance on the last character.
<nigel> .. You add tts:ipd to that span whose value is less than the sum of the advances of all the
<nigel> .. characters in that span.
<nigel> Pierre: That depends on the font, right?
<nigel> Cyril: Sure, yes.
<nigel> .. And if you combine letterSpacing and ipd?
<nigel> Pierre: This might be a really good discussion to have with ARIB, if they're willing to have a
<nigel> .. technical discussion.
<nigel> Nigel: Let's log that and see if we can look at the other attributes.
<nigel> .. They are border, marquee, ruby and text-shadow.
<nigel> .. We have some of these.
<nigel> .. We certainly have textShadow, but I haven't had chance to check if it is the same.
<nigel> .. OK looking in the English translation, the ARIB version has offset-x, offset-y, blur-radius and color.
<nigel> .. Checking TTML2, it is exactly the same.
<nigel> Pierre: Good news, I guess.
<nigel> Nigel: I think so.
<nigel> .. OK let's look at the next one...
<nigel> .. marquee is probably easy: I don't think we have that in TTML2. It does what you would think,
<nigel> .. specifying text scrolling.
<nigel> .. I recall Glenn wondering if there was some use case for animation where the same
<nigel> .. effect could be synthesised but I don't think he ever demonstrated it.
<nigel> .. I see we have w3c/ttml3#2
<nigel> Pierre: It would be good to see what effect they are looking for.
<nigel> Nigel: Agreed, and also to get some usage data.
<nigel> .. There are two more: border and ruby
<nigel> .. Looking at border.
<nigel> .. In ARIB, it is < border-style>, <border-width> and <border-color>
<nigel> .. There are lots of styles, including double, groove, ridge, inset, outset, dashed, dotted, none and hidden.
<nigel> .. In TTML2 we have none, dotted, dashed, solid and double.
<nigel> .. So for this one it looks like there may be a couple of styles they have defined, but the
<nigel> .. overall syntax and semantics are similar. TTML2 has a border-radius which ARIB does not
<nigel> .. have.
<nigel> .. Let's look at ruby:
<nigel> .. The ARIB ruby attribute is an xml:id of the content which is subject to the ruby being
<nigel> .. defined, presumably in a span with the ruby attribute.
<nigel> .. This is a simpler model than in TTML2, but does not describe alignment, position or reserved space.
<nigel> Pierre: I'm fairly confident that TTML2 does a really good job with ruby, modulo the hanging
<nigel> .. issues, the complex ones. Overall I think there should be no deltas there.
<nigel> Nigel: There are deltas, both in syntax and in functionality, but on this one TTML2 has more functionality.
<nigel> Pierre: I would think that TTML2 is a superset.
<nigel> Nigel: The one thing we don't have is infinite levels of nesting of ruby, in TTML2.
<nigel> .. ARIB's solution does support that.
<nigel> Pierre: I think there's an issue for that. We might want to ask if nested rubys are useful
<nigel> .. in subtitles and captions.
<nigel> Nigel: Yes.
<nigel> .. I think we've iterated through all the ARIB attributes now.
<nigel> .. For this issue, I think we can describe the overlaps and gaps in functionality but really
<nigel> .. what we're interested in asking ARIB is about the usage.
<nigel> Pierre: Yes
<nigel> SUMMARY: There are some overlaps and some gaps between the ARIB-TT styling attributes and those in TTML2. We are interested to know more about the usage, especially where there are gaps in TTML2.