Broadcast
The scale and complexity of producing different versions of
a show often flies under the radar, yet the process can be critical for
maximising international sales.
Content localisation – the process of creating compliant and
culturally appropriate versions for inter national markets – is growing in
complexity and volume.
The Media & Entertainment Services Alliance (MESA),
which organises a Localisation Council of content owners, distributors and
vendors, calculates that the market for services in Europe and the Middle East
generated $2bn (£1.5bn) last year. This is expected to grow by 10% annually to
hit $2.5bn (£1.9bn) by 2020.
“Localisation is the biggest single cost in our global
distribution process and is, in many ways, the most complex,” says
Brij Sharma, head of supply chain strategy and localisation at BBC
Worldwide.
“Volumes have increased significantly over the past three
years and continue to do so in emerging markets such as China and India.
For example, a new language for a single channel fully localised can represent
an increase in volume of around 850 new hours per year.”
In part a function of volume, complexity is also increasing
as content owners look to deliver their programmes faster, in more languages,
to different platforms. Localisation doesn’t merely create a new text or audio
file. In order to comply with local rules, there’s a need to create new
editorial versions, which are often delivered at a later date.
According to Sharma, this adds further complexity in terms
of management across technical, commercial, legal, financial and creative
teams, and in terms of keeping track of content versions.
“Knowing how multiple content elements for a given programme
are related to each other is an increasing area of focus, pushing us to
challenge the way we do things and to look at creative ways to use existing
resources and new technologies.”
There is an effort among specialist suppliers, supported by
MESA, to raise the profile of what has traditionally been bracketed as ‘sub and
dub’ even among broadcaster and studio personnel.
A sales exec might offer a new series at Mip to a
distributor and sign off on delivery in 17 languages by the end of the next
week without understanding the pressure on turnaround times this creates.
“It’s not just taking a programme and slapping a language on
it,” says TVT chief executive Ian Brotherston. “There are different technical
standards ranging from pre- and post-watershed versions to removal of product
placement.
”If you manage, say, The Walking Dead [which TVT
handles for AMC], you need a downstream understanding of what is appropriate in
each market to maximise worldwide sales for the content owner.”
Julian Day, business development director at Zoo Digital,
says online platforms have raised the bar in terms of quality, even for
catalogue sales.
“Previously, broadcasters were not paying much for
subtitling and it was often the last thing they’d think about. Now, when
it comes to licensing titles abroad, they find the original subtitles aren’t
good enough. They have to recreate or conform them and get them up to spec.”
Dubbing accounts for the largest portion of revenues – MESA
puts the figure at 70% – in this sector. “Experienced linguists and subtitlers
will know how to create a lip-sync dubbing script that is as true as possible
to the original,” says Day.
Acting as well as language skills are required to translate
dialogue with the same nuances as the original. There’s a “significant
shortage” of such skills in some markets, according to MESA, exacerbated by
voice talent being tied into long-term contracts with suppliers or shows.
“It’s a major concern,” says executive director Jim Bottoms.
“Given the growth in demand, content producers and their supply partners are
keen to expand the talent pool to avoid serious capacity shortages.”
Zoo Digital has tried to address this with an online
cloud-based service that aims to give casting directors access to more voice
artists globally. Live video direction of dubbing sessions and real-time
collaborative script review is also carried out online.
The method of translation is key in versioning. Whether to
meet local language needs and expectations via voice over, lip-sync dubbing or
subtitling is based on local cultural norms and regulations.
Operator brand identity may necessitate re-scripting for
voiceovers. Some channels rescript and re-voice the narration to match the
channel personality and house style – even in the same language.
“For example, two of our clients broadcast Ice Road
Truckers, a show about drivers who take on some of the most dangerous frozen
routes in the world,” says Andy Steele, vice-president of operations at TVT.
“One wants just the original US version – Americanisms and
all – while the other wants a distinct UK version, with the narration done in a
British accent, using terms the local audience would use.”
Another aspect is ensuring content is fit for the target
platform. This includes viewing it then marking changes based on factors such
as duration, what’s going on in each scene, age rating and local regulations,
which can cover sex, violence, bad language, product placement and religion.
“The process must also ensure regulatory compliance for
audio loudness and flash patterns or certain instances of varying luminance
that can trigger epilepsy,” says Brotherston.
“All these guidelines need to be codified for each piece of
content, depending on country and platform of transmission, along with the
broadcasters’ own brand values. Time of day is also important.”
A+E Networks’ docudrama required compliance and
versioning for European markets including the UK, the Netherlands,
Poland, Russia and Croatia.
A key part of the brief for TVT was to create softer
versions for daytime slots while maintaining the harder-hitting
versions for a post-watershed audience.
This meant dealing with the extreme violence and gore
depicted in many parts of the series – which is intended to reflect
historical accuracy in the dramatisation sections – while remaining
sensitive to the tension, drama and flow of the storytelling.
Special care was taken to deal with violence against
children, which is depicted in several scenes in Barbarians Rising.
In preparing the daytime versions, TVT editors made cuts just before
or in the midst of very graphic scenes – for instance, retaining a
sword combat visual right up to the moment a blade has struck a body,
but before the spray of blood.
While photosensitive epilepsy (PSE) triggers usually
mean strobe lighting or flash photography, in this case, variations
between light and shade, and the glint of steel weapons, had to be
monitored and dealt with.
Failure to pick up on even seemingly small taboos and rules
can have a high price. TVT gives examples such as allowing use of the word
‘prick’ pre-watershed in Ireland, where research shows it is deemed more
offensive than ‘fuck’; showing a tattoo on screen in South Korea; using
‘God’ as an exclamation in Poland; allowing US-style product placement in the
more restrictive European market; or depicting conspicuous drug use for a
broadcast in Japan.
Once the issues are identified, craft editors adjust the
content while retaining production values. “The content needs to remain in
keeping with the programme aesthetic and work for all the audio and subtitle
language translations,” says Brotherston. “This requires continuity of skills
and understanding of the content to make sure that edits don’t damage the storyline.”
Masks, cover shots (programme b-roll) and colour correction
can be used to retain the integrity of content and voiceover.
Additional edits of on-screen graphics and captions to
localise them for non-English-speaking markets may be required, as well as a
credit for the localisation performers.
Increasingly, platforms are requesting ‘enhanced versions’,
with extra scenes or graphics. “Linear viewers might be attracted to the
enhanced OTT versions of their favourite show, giving content owners a second
bite at the cherry,” says Brotherston.
“This generates additional complexity in version IDs and
rights management, and can lead to content having a much longer tail [requiring
it to go into longer-term archives with multiple versions that require
additional storage systems].”
While compliance editing will generally involve removal of
content, some firms take picture localisation to extraordinary
lengths. Pixar Animation created 7,482 new shots for international
versions of Cars 3, making artistic changes to frames or whole scenes.
For example, substituting language-appropriate text in
newspaper headlines to ensure a joke or plot line is not missed.
For a scene in Inside Out featuring a child refusing to eat
broccoli, Pixar changed the vegetable to peppers for distribution
in Japan. “Everyone eats broccoli in Japan. But they hate green peppers,”
says Eric Pearson, home entertainment supervisor at Pixar.
Underpinning the whole chain is a minutely detailed tracking
and reporting process. Once manually administered on spreadsheets, it’s
now increasingly part of a media asset management (MAM) system. For increased
flexibility and ease of access, vendors like TVT and Zoo Digital have also
centralised their MAM in the cloud.
With such complexity, it would be a logical step to
introduce automation to streamline services. Indeed, some proponents of machine
learning suggest that incumbent service providers are reluctant to introduce
new technology for fear of damaging their revenue.
“For the same amount of money you spend on a transcription
house, you can use machine learning today to deliver speech to text,
localisation and an almost infinite other variety of data tasks,” says Josh
Wiggins, chief commercial officer at metadata solutions provider GreyMeta.
“At the same time, you end up with richer content because of
the comprehensive and detailed metadata that results from applying artificial
intelligence.”
However, automatic transcription is still a long way from
converting a soundtrack into accurate text in real-time. According to
subtitling specialist Screen Systems, the challenges include sound effects and
background music, different voices and accents – and switching between them –
unexpected vocabulary and punctuation.
War For The Planet Of
The Apes
Digital Media Services (DMS) was responsible for
creating 500 promotional assets, including localised TV and social
media spots, digital billboards and a variety of online content to
support worldwide distribution of Fox’s theatrical release War For
The Planet Of The Apes.
To work on the assets, DMS received master files from the
Fox Media Cloud system and uploaded them to its own platform,
Panther Online.
“Requirements vary considerably, with some countries
requiring one or two traditional spots to support their marketing,
while other countries order 20 and even up to 40 bespoke variations,”
says Michael Petch, director of customer services.
Many of the briefs only specified the text needed for
the call to action details per asset, leaving DMS responsible for
incorporating local specifics such as voiceover, age ratings, URL and
social media tags.
Voice artists recorded the localised voiceovers at ADR
facilities. In most instances, content was also dubbed and subtitled,
while graphics cards were created to match the quality of the
studio’s masters. Locallanguage-speaking customer service co-ordinators
checked every asset before sending to Fox for sign-off.
A glaring illustration of this occurred when Newcastle
United was referred to as ‘the black and white scum’ in an automated
mistranslation of a presenter’s words during Match Of The Day 2 in September.
“AI translation technologies have improved greatly in the
past few years… [but] we do not believe that the toolsets on offer today can
replace human input when it comes to adapting long-form content such as drama,”
says Sharma. “In our experience to date, it’s often quicker to start a
translation from scratch than to start from whatever base a machine might
create.”
Zoo Digital’s Day suggests that a good translator can accept
or edit suggestions from a machine, “but it’s a supporting tool at this stage.
Ultimately, we still rely on a human check.”
Pinewood, which creates formats for theatrical content, is
investigating the use of language-specific speech recognition as an extra step
in its QC workflows, but says reliability is an issue – especially when it
needs the tools to recognise 44 languages.
TVT has speech and image recognition software in beta. “We
are starting to use AI from the EDLs and metadata we’ve previously created,”
says Brotherston.
“Image recognition might be useful in a future editorial
process – for example, if there’s a real-life plane crash we could search
programming for incidents of plane crashes and alert broadcasters, or reedit
and redistribute the video as needs be.”
Benefits from automation are most likely to be derived in
workflow management and orchestration, to reduce duplication.
“Each piece of content is in effect a project, with its own
sets of rules around which version to create, what voices to cast, approvals,
merchandising, rights and so on,” says Sharma. “This is highly labour
intensive, so a good candidate for automation.”
No comments:
Post a Comment