Report - Subtitling- 3D and Timed text
The UK Section first meeting outside London included the world’s first demonstration of DVB subtitles (EN 300 743) in Stereoscopic 3D as well as a stunning exposition on Timed text whether for distribution as in SMPTE ST2052 or EBU –TT as an ”authoring archives and transfer format”.
About 20 people heard Peter Weitzel, Hon secretary of the UK section, point out that only about one third of the 200 Members of Section live in London , and how the Board of Managers had decided on holding meetings and events outside London whether there were “clusters” of membership and interests in our industries.
The Regional seminar had been held in at MediaCity: UK in Salford, and there were two meetings – events where members and other can informally discuss and then hear speakers cover aspects of current concern- here in East of England – marking an arc from the University of Essex at Colchester through Ipswich Cambridge to Ely and a more circular cluster in the South of England around Winchester Basingstoke and Southampton in May.
Although this it is first year, a tradition had been set that the UK section meets where members work – but for a firm like Screen Subtitling – do not have large demonstration rooms – so we were meeting in the Hotel at the end of the road from their HQ - a Historic Rectory which has been their home since 1976.
The Topic of subtitling was something where the UK excelled in having a number of equipment manufacturers and Broadcasters who know that access services enhance the viewing experience for an audience beyond those who acknowledge their they have a disability – and he said that he was at a demonstration the following day organised by the Cinema Exhibitors Association – looking at systems for theatric use. With the BBC subtitling 100% of its output (a voluntary decision) and all other broadcasters exceeding their regulatory quotas, it was important that SMPTE looks to the future - and this was the purpose of this meeting- Firstly Subtitles/Captions in a 3D scene, and secondly the use of timed text for subtitles.
He had a personal interest in this having 15 years ago been the major author of the DVB-ST specification EN 300-743– and then under the auspices of the UK Digital Television Group he chaired an ad hoc group to replace the EBU 3264 file format - it failed – the technology of XML was not there yet.!
Simon Hailes CTO of Screen started with a review of subtitling (captioning) for 3D content – and how the subtitle need to be read but not seen. That is they should be part of the picture and not intrude – and the greatest requirement of subtitling in 3D was that the subtitles were set to the same vergence and hence depth position as the speaker they referred to and that they did not break the 3D effect through conflicting with other elements in the image. This gave a very different on screen presentation – with “speech bubbles” following the speaker rather than a “lower third super” over the whole picture.
But he pointed out that there were a lot of important details that need to be included in a specification for 3D subtitles – from the obvious need for a clear font (perhaps at a smaller line size – but it is on a bigger screen). The key factor was the positioning – either in a prepared situation – or in real time by an automated process which set the depth …. But this needs to delay the video so that the subtitles can be moved as the scene changes – and often moving the subtitle before the scene moves is a good technique.
He emphasised the need for sub pixel disparity positioning and of course the basics static fine x/y positioning of the subtitle/caption. In 3D it is even more important that what you author is what the viewer sees- i.e. the workflow through to final rendering has to be consistent, and also there may be a need to resize the caption an it retreats or advances in depth. .
He pointed out that any specification should be non-ambiguous and led to a consistency of implementation
He reflected that the DVB subtitle specification EN 300 743 has bit mapped images and thus are created/ rendered at the broadcasters site rather than by conversions/ rendering of a text string in the TV/ STB as happens in Teletext, EIA 608 or 708 and Timed text systems like SMPTE2052
The DVB subtitle specification now includes 3D features and he noted that it includes Fixed and variable disparity , nonlinear movement with subs pixel disparity, and a few “polishing up” enhancements to the existing specification But unfortunately it did not achieve scaling are sub pixel static positioning .. And with the advent of higher frame rates raises questions on how some phrases are to be interpreted.
But 3D subtitles for TV are here - and after the presentations there was a demonstration of s Sony TV decoding a 3D DVB subtitling stream (delivered as a MPEG transport stream) … the first demonstrations of this in the world. Members commented on how it looked better than caption rendered by a character generator burnt in on the video.
John Birch Strategic Partnership manager at Screen who has been a leading contributor to the SMPTE and EBU time text standards introduced the basics of timed text - Text, timing information plus style and metadata usually expressed in XML. He pointed out that work on Timed Text started just over 10 years ago when the TTML Working Group of W3C started work on TTML DFXPand its authoring counterpart AFXP. But it was not until October 2005 that the first draft was available – and the W3C Recommendation was released on 18 Novembers 2010 ... this delay had caused some who could not wait to form their own standards – like SMPTE ST 428-7 Subtitle for Digital Cinema.
He then pointed out that profiles within TTML were not constraints or subsets within the TTML Specification but Profiles allow browsers / processors to determine if they are able to display or process a document correctly.
There were three main profiles:
Transformation Profile: intended to be used to express minimum compliance for transformation processing.
Presentation Profile: intended to be used to express minimum compliance for presentation processing
Full Profile: intended to be used to express maximum compliance for both transformation and presentation processing.
Looking towards subtitles and captions the TTML Simple Delivery Profile for Closed Captions (US) defines an interoperable delivery profile using the core features specifically designed for online presentation. The profile allows the community of TTML users to construct constrained presentation engines for TTML for captioning.
SMPTE ST 2052-1:2010: defines the SMPTE profile of W3C Timed Text Markup Language (TTML), which may represent Captions or Subtitles. It identifies the features from TTML required for interoperability between display systems for the format. SMPTE-TT also defines some standard metadata terms to be used, and adds some extension features not found in TTML.
Birch pointed out that SMPTE2052 was for the display of subtitles whilst the other Subtitle form of Timed Text EBU-TT was authoring and archiving format used by the broadcaster. It replaces EBU3264 “STL files” which dates from 1989 and is almost universal use in Europe for both Teletext and Open subtitles/Captions. EBU-TT also defines restrictions on the use of TTML-DFXP vocabulary and grammar.
He then showed that the SMPTE and EBU Timed text Captions/ Subtitles lay alongside others derived from W3C TT - both Flash and ITunes use a timed text delivery. Mention was made that the EBU was moving onto a live format of Subtitles replacing the existing Proprietary NEWFOR serial line protocol.
Birch then took a deeper look at each of the formats and outlines the structure of the specifications In particular pointing out the “-0 “part of the SMPTE specification which gives a High level over view and then explaining in some detail how EIA 608 Subtitles can be Tunnelled through the SMPTE TT file (ST2052-10-2012) and how work was in progress to do the same with EIA 708 and ST 428 Subtitles. He pointed out that the FCC had said that SMPTE 2052 was “safe harbour” to comply with the Twenty First Century Communications and Accessibility Act 2010. And that from 30 March2013 not only Pre-recorded TV material but now “Live or near live” TV material on the web has to have captions of equal or better quality when distributed online.
Turning to the authoring format EBU TT – that had a number of parts – including how to incorporated legacy EBU 3264 STL within EBU-TT and a extension to have a single EBU TT-4 Annotated file from which a range of EBU—T-1 files can be formed depending on the use case . He showed the structure of typical EBU-TT files showing the tags which were Metadata, style timing and the actual subtitle text itself. He pointed out that these would be interpreted and rendered by a DVB –ST coder or translated into Text conveyed subtitles/captions - Teletext ,EIA607 or 708 and Timed text- for rendering in the receiver.
Birch in closing outlined the decisions that needed to be made as you decide which format to use and emphasised that both EBU and SMPTE Timed Text were very close to being mature and with multiple implementations but pointed out that if the subtitles/Captions were not in the” master “(IMF or similar) it was not easy to add it later!.
There was good conversation over refreshments while demonstrations of the 3D DVB subtitles were being shown.
This was a very successful first meeting of the UK Section outside London – and much talk was of the next meeting of members in the Eastern Counties.
The UK Section thanks Screen for providing the venue, speakers and demonstrations and to Pixelpower who sponsored the ample refreshments.
- All posts: