Conference Report: MESA Europe Forum

Chief Nimdzi Researcher attended the MESA Europe Forum in London. This is an exciting event that is unfortunately overlooked by many in the localization industry. If you did not attend, don’t worry. You can read NImdzi’s report and maybe we’ll see you there next year!

Have you ever wanted to localize Game of Thrones, Friends, or Mad Men? Now would be a good time to start. Media localization is not only glamorous, it’s arguably the most attractive segment to be in the language services market right now.

The film & TV industry is undergoing a transformation, and under the skin of the chrysalis, the amount of video content is growing fast. On American TV, the number of original scripted programs has increased 2.5 times in the last 10 years, from 192 titles to 487. Throw in the original series made-for-web platforms and private channels, and you get a deluge of content. Netflix and its rival OTT providers have been aggressively pursuing international markets and rolling out original content. Traditional broadcasters are on the move to achieve digitalization as their traditional linear revenues are declining. They realize that in the next generation, there will be no broadcast TV, only the likes of Netflix: on-demand, personalized, programmatic, with original content in many languages.

Mid-sized Netflix localization vendors are literally exploding in revenue. Voice and Script International grew 88% last year to almost USD 40 million, Zoo Digital has seen an increase of 143% over two years to USD 28 million. The two of them together became fastest growing LSPs in 2017.

Record growth, a siloed niche cultivated by a handful of vendors, and more volume to come – reflect on this for a moment, and then if you’re even a little bit entrepreneurial, you’ll want to get into media localization.

Check out the photos from the event by viewing the full album here 

The conference

 With these thoughts in mind, Nimdzi embarked on the flight to snow-blanketed London, where the European branch of the Media and Entertainment Service Alliance held its annual forum. If you’ve never heard of MESA before, it’s best to consider this an industry association. The European branch has about 45 corporate members – it’s a spinoff of the US-based organization that has been going on for 10 years.

On the outside, the “Content Workflow Management Forum” felt just like a regular mid-sized localization trade show:

  • About 170 participants
  • A single day and a single track
  • Lots of networking and people-to-people business making

But there was one exception – there weren’t too many familiar faces around. MESA is like a well-kept secret bottle of wine – few LSPs or tech vendors know about it, so only the select connoisseurs can benefit.

Walt Disney, Discovery Channel, BBC, Deutsche Welle, Twentieth Century Fox Film, Warner Brothers, Viacom, and even the Pokemon Company – brands one has seen on screen since childhood represented the buyer side. Services vendors brought in sizable teams of 5 – 7 people, with the largest delegations from BTI Studios, SDI Media, Zoo Digital, Deluxe, and IYUNO.

To us, all of this was new. Despite having attended close to a hundred localization conferences before, I’ve met only one person whom I knew from before, which shows how insular media localization is.

The topics

Most discussions at the conference revolved around digitalization, automation, metadata, and new regulations such as mandatory audio descriptions for movies. Unlike translation conferences, not even once did the media vendors complain about price pressure, low rates per word, or declining margins. The talks focused on the immediate challenges rather than the good old past or the looming robotic future.

Addressing growth

Today’s increase in original content means more dubbing and subtitling work in the pipeline, and in the words of Julian Day of Zoo Digital, “you will need a bigger boat.” The challenge for the buyers (or “content owners” in the professional argot) is to scale up production in a smart way without simply throwing more human bodies at it.

Internal localization teams need to get the localized versions to their distribution companies on schedule in the environment where localization is often an afterthought. At a client panel, Jan-Hendrik Hein of A+E Networks explained that the pressure on turnaround times is mounting, and in these conditions, buyers are open to talking to vendors that bring new ideas to the table.

Vendors are looking to address the challenge with software: not spreadsheets anymore, but cloud-based systems with automated workflows, connected workspaces, asset management functionality, and early AI features. Solutions range from adapted translation workflows in XTRF (Haymillian) to OOYALA, a large Silicon Valley start-up that can automatically detect celebrities via face recognition and tag videos with metadata for personalization and advertising.

Format and medialoc process standardization

One of the topics that interested content owners a lot was standardization. Media content is meant for many screens: from a mobile display to a cinema theatre, and due to the variety of formats, localization teams sometimes receive up to 8 versions of original content, which then might multiply by the number of languages, resulting in a database of content that is challenging to manage. Large files for 4k and 8k videos and videos optimized for HDR10 can also become unwieldy to store and transfer.

Even film titles are still not standardized. MESA Forum featured a presentation by Will Kreth who leads EIDR, an initiative to assign a unique id to each movie. With 2 million titles in the database, the work is only starting.

Furthermore, there is work underway to publish modern guidelines to media loc workflows, as well as dubbing and subtitling practices. A handbook by Yota Georgakopoulou from Deluxe Media is scheduled to come out this autumn. However, there was no talk of an ISO standard in this field.

Metadata for TV and film content

Tagging videos with as much metadata as possible is essential to personalize them and improve consumption, and it is even more vital for monetization via real-time ads tailored to what viewers see on the screen. New software platforms can do automated content enrichment, as well as detect characters and celebrities. But the challenge remains for content owners with large archives to do the manual part of the tagging. They need to create internal metadata teams, develop taxonomies, and assign tags to each video. The taxonomies are not standardized across the industry, so careful planning and research is required. While crowdsourcing video tagging is an option for some media archives, I believe there is business potential around professional metadata services.

Machine learning applications: automated translation and speech-to-text

Media localization professionals are exploring applications for machine learning:

  • For example, technology based on deep learning highlights changes in a new cut of the movie, including the extended editions and directors’ cuts. It can expose the new content only for localization, making it much easier to deliver the localized version on a tight schedule.
  • Character face and voice recognition can help with exact time stamps for dubbing voice actors, and it also isolates their character’s words for them so that it is easier to record every phrase in one go.
  • Bots also help detect nudity and profane language to limit exposure to underage viewers.

Neural machine translation is not yet good enough for theatrical releases, but together with speech-to-text technology, it can automatically create subtitles for fast-moving content such as news, conference speeches, and eLearning. With automated transcription and post-editing, it can reduce the workload tremendously. Deutsche Welle innovation manager Peggy van der Kreeft explained that it can cut one day of subtitling to an hour or two.

SUMMA – BBC/DW platform for news monitoring

One project that exemplifies what can be done with modern language technology applied to an information business is SUMMA project – a platform for large-scale media monitoring. It is developed under the EU Horizon 2020 programme and will be primarily implemented in BBC and Deutsche Welle. Showcased by Peggy van der Kreeft, it translates news from 30 countries into one stream of English language summary stories extensively tagged with mentioned personalities and organizations. Journalists working on new stories can use the platform to quickly get a grasp of everything said on their chosen topic. Behind the scenes, this piece of technology directly out of a James Bond/Batman movie uses speech-to-text, entity extraction, auto-summarization, and machine translation components. No human translation yet – there is simply too much content to afford it.

The next MESA workshop will take place in Paris on March 20.

Opening img: squaremeal.co.uk


Stay up to date as Nimdzi publishes new insights.
We will keep you posted as each new report is published so that you are sure not to miss anything.