ADOPTING NEW SUBTITLE FORMATS TO MEET AUDIENCE NEEDS new... · edited to create subtitle blocks (4). The Centre for Deaf Studies in Bristol reviewed the research on television subtitling

ADOPTING NEW SUBTITLE FORMATS TO MEET AUDIENCE NEEDS

N I G E L M E G I T T , I R T S U B T E C H 1 S Y M P O S I U M

2 5 M A Y 2 0 1 8

2

A BIT ABOUT THE PRESENTER

NIGEL MEGITT

Executive Product Manager, Design + Engineering, BBC

Other roles:

Co-Chair, W3C Timed Text Working Group Co-Chair, EBU Timed Text Group

Contributor to many others…


3 ADOPTING NEW SUBTITLE FORMATS TO MEET AUDIENCE NEEDS

A BRIEF HISTORY OF AUDIENCE EXPECTATIONS

4

2

2.1 Research & Regulation

The original guidelines for subtitling in the UK were informed by research carried out in the late 1970s by a team based at Southampton University on behalf of the Independent Broadcasting Authority (IBA). This pioneering work was difficult because of the nature of available television equipment at the time and subtitled television was a novel experience for the participants (2, 3). The guidelines were published in 1982 and contained extensive guidance on how scripts should be edited to create subtitle blocks (4).

The Centre for Deaf Studies in Bristol reviewed the research on television subtitling in a report for the BBC and Independent Television Commission (ITC) in 1992. The report raised concerns about the existing research and the lack of follow up work in several areas including the issue of whether subtitles should be edited or presented verbatim (5). In 1996 the ITC commissioned research on viewers' preference for block subtitles or scrolling subtitles for news coverage. The report recorded no strong preference for block or scrolling subtitles but argued for edited subtitles, despite a majority of respondents preferring verbatim subtitles. It also highlighted problems with subtitles being delayed and obscuring other information (6). The ITC published new guidelines in 1997 that were updated in 1999 to include the new digital television services and the move to DVB subtitles. However, apart from specifying the Tiresias font the approach replicated the existing Teletext delivery (7).

In 1999 a book was published that attempted "to establish the common ground between intra- and inter-lingual subtitling." Based on research at the Centre for Deaf Studies the book deliberately conflated subtitles for access services with subtitling for translation (8). It seems to mark the start of the field of Audio Visual Translation where arts academics have focused on the subtitler and have been dismissive of user preferences, even framing them as "political" (9) or dismissing them as "a huge mistake" (10).

Since Ofcom took over from the ITC they have commissioned research on subtitling speed (11) with a view to revising the ITC guidelines, which they published following a public consultation in 2006. These new guidelines are considerably shorter than the ITC document, removing the guidance on editing and formatting of subtitles. However, they still contain some legacy issues inherited from Teletext (12). Since 2013 Ofcom has been focusing on a programme of work measuring the quality of live subtitling (13).

2.2 The Changing Media Landscape

In 2009 BBC Online published a set of guidelines outlining the subtitling requirements for AV content presented on the bbc.co.uk web site (14). These were based on existing television subtitle guidelines and existing research (5, 6, 15). We have been reviewing the validity of this guidance for the new viewing platforms and viewing contexts. With television programmes available on computers and portable devices the context for television viewing is a more individual experience and increasingly portable. In fact, by the end of 2013 viewing of the BBC's iPlayer on tablets overtook viewing on computers (16).

As we reviewed the BBC Online guidelines we also found problems with the subtitling guidelines for television and the persistence of legacy constraints. The Teletext display grid of 24 rows by 40 columns was designed to for CRT displays with interlaced scanning. The font was fixed-width and a double height font was used for subtitles. This layout has been used to conduct most research over the past 30 years. In the UK the format for subtitle delivery to the home has moved on. Since October 2012, most UK broadcasters deliver subtitles in DVB format and iPlayer uses the TTML format.

Figure 1 – Teletext subtitles on a CRT TV

Around 40 years ago, the BBC began broadcasting subtitles using the excellent new Teletext system. Electronic text displays were basic, and this system provided readable text in a small number of colours with some positioning. Great! Over the years broadcasters like the BBC integrated the Teletext into their workflows, standardising on storage formats (STL), and using ad hoc systems for inserting live subtitles (e.g. Nufor), and specifying how to carry the teletext in scanning video streams in ancillary data sections (SDI).


A BRIEF HISTORY OF AUDIENCE EXPECTATIONS SUBTITLES: THE EARLY DAYS

5

… the audience began to get used to computers that could produce nice looking text at home. The old monospaced Teletext font began to look a bit dated. DVB created a bitmap specification that encoders could generate from the Teletext source data, to make the text look nicer. Some platforms like Sky rendered the Teletext in the client device. The common backbone to the workflow of Teletext remained though.


A BRIEF HISTORY OF AUDIENCE EXPECTATIONS MEANWHILE…

6

Teletext was great, but in 2018 it doesn’t quite look fit for purpose. It can’t do some things that we need for global use – on the right are just a few. These are things that the web can do that the audience now just expects. Your phone can do these! More importantly, they are necessary for making video accessible.


A BRIEF HISTORY OF AUDIENCE EXPECTATIONS BEGINNING TO LOOK A BIT WRONG?

Display Unicode characters like €, ♫, 🤔

Use a wide variety of colours

Different fonts, including proportionally spaced

Precise positioning

Begin at the left edge

Handle bidirectional text שלום

Or vertical, or Ruby…

Work nicely on the web

Carry metadata


BROADCASTER INFRASTRUCTURE

8


BBC SUBTITLE WORKFLOWS – CURRENT (2018) Subtitle inserter

Prepared workflow

Live workflow

Shortform (iPlayer exclusive, clips)

Conversion (major)

Processing (minor)

Subtitler(Red Bee) Playout

Broadcast Coding &

MultiplexingTV

BBC Standard Media PlayerBBC R&D

EBU-TT with embedded STL

DVB Bitmap & Teletext

Media Encode and Package

services

STL

Teletext VANC (2)

Teletext (VANC)

“TTML” (3)

STL

NUFOR(1)

Subtitler(BBC?) EBU-TT-D (clips etc)

Internal Web production/CMS tools

EBU-TT-D

Digital archive

File delivery receiving

system

EBU-TT-D (live to VOD)EBU-TT with

embedded STL

6 formats, 3 conversion points.


9


BBC SUBTITLE WORKFLOWS - VISION Subtitle inserter

Prepared workflow

Live workflow

Shortform (iPlayer exclusive, clips)

Conversion (major)

Processing (minor)

2 file formats, 1 conversion point.


Subtitler(Red Bee) Playout

Broadcast Coding &

MultiplexingTV

BBC Standard Media Player,Freeview Play

Digital archive

DVB Bitmap & Teletext

& DVB TTML

Media Encode and Package

services

EBU-TT-D (VOD)

EBU-TT (prepared)

Subtitler(BBC?) EBU-TT-D

Internal Web production/CMS tools

File delivery receiving

systemEBU-TT

EBU-TT

EBU-TT pt 3

EBU-TT pt 3

EBU-TT-D

EBU-TT-D

EBU-TT (live captured)

EBU-TT-D in DASH (stream)

Archive Search

10

The broadcast industry seems to be heading strategically towards: •  IMF for mastering and archive •  IP streams (e.g. SMPTE 2110) for playout and live

This means that we are generally heading towards a de-embedded future, where subtitles are not embedded directly into other media. But there may be a case for doing that with e.g. MXF deliverables intended for playout. This doesn’t change the vision for subtitles, but it might have a big impact on how subtitle streams are carried and how the audio and video are managed.

Everything is just an object. Subtitles were objects first!

LONGER TERM

IP EVERYWHERE?


11

Quite a lot of Teletext-based solutions depend on physical hardware, for example to insert subtitles into an SDI stream. We are moving more and more towards cloud based solutions, especially for providing web-based streams. We just can not spin up and spin down processing instances when there’s a dependency on a limited number of physical machines. Whatever solution we choose needs to be software and IP network based so we can choose the right deployment model.

LONGER TERM

THE CLOUD!



ADOPTING NEW STANDARDS

13

There are a lot of subtitle formats (not so many standards)! BBC prefers: •  Open standards – freely available, developed in an open

process •  As few standards as possible, or minimal transcode

requirements •  Technology that fits business processes The standard needs to support: •  Everything Teletext can do, and the things it can’t do •  Prepared subtitles •  Live subtitles •  Hard of hearing and translation •  Broadcast and web distribution and playback •  Support for the whole broadcast workflow, i.e. the right

timing and supporting metadata.

Our preferred choice is the TTML family. We helped make it, initiating the work in W3C back in 2003, and have worked with W3C and EBU since to create profiles that meet our needs, and the needs of our audience. TTML profiles include EBU-TT, EBU-TT Live, SMPTE-TT, IMSC, ARIB-TT etc. Industry seems to be converging on TTML globally: •  IMSC in MPEG CMAF and IMF (even on iOS!) •  EBU-TT-D and IMSC in DVB TTML •  EBU-TT-D in HbbTV 2.0, Freeview Play •  IMSC in ATSC 3.0

ADOPTING NEW STANDARDS

WHICH STANDARDS?


14

Mostly no, the standards are in a pretty good state, and have maintenance routes. One missing area is for subtitles in IP infrastructure: •  EBU-TT Live in SMPTE 2110? That’s good for de-embedded workflows, but another where questions are often asked is where subtitles are embedded into AV assets:

•  TTML in MXF? Some work may be needed, not sure. •  EBU-TT Live in SDI? Might be a short term gain, possibly

not worth it if we’re going straight to SMPTE-2110.

ADOPTING NEW STANDARDS DO WE NEED MORE STANDARDS SUPPORT?


15

Workflow step What we want Can we use?

Authoring EBU-TT + EBU-TT Live OK

Archive and Exchange EBU-TT OK

Playout EBU-TT + EBU-TT Live No support

Encoding for broadcast EBU-TT Live -> multiple No commercial options

Broadcast Distribution EBU-TT-D/IMSC Poor support?

Broadcast Player EBU-TT-D/IMSC OK + more coming

Online Distribution EBU-TT-D/IMSC OK

Online Player EBU-TT-D/IMSC OK

ADOPTING NEW STANDARDS STEPS ALONG THE WAY


16

Broadcast infrastructure seems to cost a lot and be refreshed as rarely as the business thinks it can get away with. There may in the future be disruptors that offer new cheaper ways to implement broadcast workflows. Assuming there are not, we will need to work with our major suppliers to make sure any new functionality is either built into existing equipment or included in any technical refresh projects. Often there is a “chicken-and-egg” problem! Result: likely to take years rather than months.

ADOPTING NEW STANDARDS HOW LONG WILL IT TAKE?



CONCLUSIONS

18

We know the engineering problems we want to solve.

We know how we want to solve them.

The technical standards are mostly in place.

There is momentum towards convergence in the industry.

We have more work to do specifically in automated playout and in encoders and packagers.

Calls to action: •  if you’re buying new kit, consider

moving to new standards. •  If you’re selling kit, put this on your

development roadmap. •  If you’re representing the audience, let

your broadcasters (and maybe even regulator) know what you would like to see and what editorial proposition you would like.

CONCLUSIONS WE’RE ON THE WAY


D E+ THANK YOU! NIGEL MEGITT [email protected]

20

ONWARD…

MORE INFORMATION

BBC Subtitle Guidelines: http://bbc.github.io/subtitle-guidelines/

BBC Academy Guide “How to create subtitles”:

http://www.bbc.co.uk/guides/zmgnng8

BBC R&D publications on accessibility: https://www.bbc.co.uk/rd/topics/accessibility


Documents

ADOPTING NEW SUBTITLE FORMATS TO MEET AUDIENCE NEEDS new... · edited to create subtitle blocks (4). The Centre for Deaf Studies in Bristol reviewed the research on television subtitling