W3C home > Mailing lists > Public > public-tt@w3.org > May 2016

Re: Examples for TTML 2 WebVTT Mapping

From: Andreas Tai <tai@irt.de>
Date: Mon, 9 May 2016 15:51:17 +0200
To: public-tt@w3.org
Cc: David Singer <singer@apple.com>, Simon Pieters <simonp@opera.com>
Message-ID: <573095D5.6080103@irt.de>
Dear all,

I added now an explanation of the folder structure 
To make the access of the html pages easier they are now also hosted on 
github pages [1][2][3][4][5].

In the following I want to share some implementation experiences.

Complexity of mapping for developers
The developer, who had implemented the mapping samples, had good 
knowledge of TTML and basic knowledge WebVTT. He noticed first that the 
mapping document alone was not sufficient to do the mapping. He needed 
to go back to the WebVTT spec to implement some of the features. One 
problem he had was that the WebVTT spec has advanced since the mapping 
document has been written. In case of a bad result it was therefore 
difficult to decide if this was an error of the mapping document or if 
the WebVTT spec has changed since publication of the mapping document 
(see for example [6][7]). Although he was quite open to the WebVTT spec 
at the beginning he found some of the principles in WebVTT hard to 
understand and also fundamentally different to TTML. At the end it 
needed around three iterations and consultation with a mapping document 
editor to come to a conforming mapping.

Preservation of information
Especially in respect to the use of regions there seems no lossless 
mapping to VTT regions. One difficulty was to test WebVTT regions 
because lack of implementation (they seem one provider that implemented 
this feature). It is not clear if styling of the region by using the 
::cue-region pseudo-element fully works. For the parts we tested 
(background color of region) we could not find an implementation to 
prove it. For the region feature and other features the preservation of 
presentation intention depends largely on conforming implementation (see 

Testing against existing WebVTT implementation
The testing of WebVTT implementation against expected rendering gave a 
very mixed result. By using the mapping strategies defined in the 
document none of the implementations we looked at provided the expected 
result. Also the feature where implementations deliver unexpected 
renderings varied across the implementations. In some cases the 
deviations have been so serious that the rendering was not acceptable 
given the original intent.

Spec Version
The mapping exercise confirmed what was suspected before: the mapping 
document faces one main problem in respect to the version of the WebVTT 
spec. It needs to take into account three versions:

a) the version that was the reference at time of writing (of the mapping 

b) the latest version of the spec

c) the version of the spec that is implemented in different clients

This task is impossible to solve at the moment. For fixing errors in the 
documents it seems best to stick with the spec version at time of 
writing because that version was seen as stable at that time and seems 
still a reference for some implementations. Changes in the WebVTT spec 
can be incorporated once the spec gets finally a stable status. It may 
also be helpful to signal which of the WebVTT features are at risk and 
which of the new edits have been implemented already.

At the moment it cannot be recommended to use the mapping document in 
operation. Apart from the fact that it is still an editor’s draft, the 
results could lead to unacceptable presentations in some environments. 
Although in theory the results could be much better this takes more 
conforming and interoperable WebVTT implementations to become an asset.

I would like to encourage members of the TTML and VTT communities to 
review the few sample mappings. If there is an issue or error please 
check first the mapping document. If it is an error in the mapping 
document please file an issue there[7]. If there is an error how the 
mapping is applied in the samples, please file an issue in the 
IRT-EBU-TT-D Application Sample repository.

Best regards,


[6] https://github.com/w3c/ttml-webvtt-mapping/issues/24
[7] https://github.com/w3c/ttml-webvtt-mapping/issues/21
[8] https://github.com/w3c/ttml-webvtt-mapping/issues

Am 05.05.2016 um 16:32 schrieb Andreas Tai:
> Dear all,
> IRT has been working on four examples to test and demonstrate the 
> TTML-to-WebVTT mapping specified in "Mapping Between TTML and WebVTT", 
> W3C Editor's Draft 16 February 2016 [1].
> The result can be found in the branch "webvtt-mapping" of the "IRT 
> EBU-TT-D Application Samples" repository on github:
> https://github.com/IRT-Open-Source/irt-ebu-tt-d-application-samples/tree/webvtt-mapping/webvtt-mapping 
> The "IRT EBU-TT-D Application Samples" are published to support 
> EBU-TT-D implementation. As EBU-TT-D is a subset of TTML 1 and IMSC 1 
> the samples can also be used for TTML and IMSC 1 implementation.
> The "Application Samples" include TTML documents and example 
> renderings as images and videos.
> Because a mapping from TTML to WebVTT based on the W3C mapping 
> guideline should deliver an equivalent result, the EBU-TT-D rendering 
> videos are also used to test the WebVTT rendering. It can be accessed 
> by hosting the webvtt [2] folder on a webserver and accessing the 
> documents in the html directory [3].
> A description of the implementation experience will be provided soon.
> Best regards,
> Andreas
> [1] http://w3c.github.io/ttml-webvtt-mapping/
> [2] 
> https://github.com/IRT-Open-Source/irt-ebu-tt-d-application-samples/tree/webvtt-mapping/webvtt-mapping
> [3] 
> https://github.com/IRT-Open-Source/irt-ebu-tt-d-application-samples/tree/webvtt-mapping/webvtt-mapping/html

Andreas Tai
Production Systems Television IRT - Institut fuer Rundfunktechnik GmbH
R&D Institute of ARD, ZDF, DRadio, ORF and SRG/SSR
Floriansmuehlstrasse 60, D-80939 Munich, Germany

Phone: +49 89 32399-389 | Fax: +49 89 32399-200
http: www.irt.de | Email: tai@irt.de

registration court&  managing director:
Munich Commercial, RegNo. B 5191
Dr. Klaus Illgner-Fehns
Received on Monday, 9 May 2016 13:52:44 UTC

This archive was generated by hypermail 2.3.1 : Thursday, 5 October 2017 18:24:28 UTC