In response to soaring content localization needs, online streaming giant Netflix launched a recruitment drive to attract fresh translation talent in March 2017. The program, named Hermes, was billed as “the first online subtitling and translation test and indexing system by a major content creator,” and was advertising the fact that “Netflix is Looking for the Best Translators Around the Globe.”
At that time, Netflix movies were being translated into more than 20 languages, and the scale of the localization was on overdrive following the launch of the service globally just a year earlier in January 2016.
By March 2018, one year after the Hermes launch, Netflix had issued a statement on its website to announce that the program was being closed. The notification read: “we have reached our capacity for each one of the language tests due to the rapid popularity and response from applicants all over the world. Therefore we are closing the platform to future testing at this time.”
At the time, Slator reached out to Netflix for comment on the closure of the platform, which seemed rather unusual. “We [don’t] have anything to add at this time outside of the messaging posted on our site,” a Netflix spokesperson told Slator in an email.
Leaving Onboarding to the Experts
Now, Netflix has provided more color on the reasons behind the closure of the Hermes project and it seems the company may have bitten off more than it could chew. In a presentation at the Languages & The Media 2018 conference in Berlin, Allison Smith, Program Manager, Localization Solutions at Netflix, explained how the Hermes project had been highly ambitious in its goal of testing, training and onboarding thousands of new translators.
Yet after much introspection, Smith said, the team pivoted and decided that those activities were better left to the ten or so localization vendors that Netflix partners with, allowing Netflix to focus on tasks more aligned to its core competencies such as content localization workflows, engineering and development. It is a tech company after all.
When asked by an audience member during the Q&A how successful Hermes had been in onboarding translators, Smith responded that it had been valuable in other ways. The project generated lots of new ideas that Netflix has taken forward such as scheduling improvements, enhanced style guides and continued development of a cloud-based content localization platform, Smith said.
“Netflix aimed to own the full process from subtitler recruitment through to working in our tools, and this started with Hermes. While we learned a lot and did get value from the test, after introspection and analyzing our core competencies, we decided vendors were better suited to use their core competencies and add value to the content localization ecosystem by owning the recruiting, training and onboarding processes.” — Allison Smith, Program Manager, Localization Solutions, Netflix
Netflix Asks the Audience
Many of Netflix’s preferred localization vendors have their own translation environments, which translators may use as opposed to Netflix’s own tool. It’s not clear how much of the translation work is being done in the Netflix platform itself and how much is being done in external platforms.
Slator Machine Translation Expert-in-the-Loop Report
Still, Netflix continues to seek feedback on its localization platform to inform the development roadmap. To collect additional on-the-spot feedback from the 370 Languages & the Media audience members, many of whom translators, a snap poll was taken during Smith’s presentation asking people what additional features they would like to see built in to Netflix’s timed-text tool. The most popular feature requests were for spell check and autocorrect capabilities, an offline version of the cloud platform, and translation memory (TM) integration.
The most popular feature requests were for spell check and autocorrect capabilities, an offline version of the cloud platform, and translation memory (TM) integration.
The same audience, professionals from all areas of the media localization industry, were also polled on how they feel about using machine translation (MT) in subtitling. The response was conservative. On a scale of one to four stars, with four being the most positive, 61% gave one or two stars. 19% gave three stars while 20% gave the full four stars.
Smith was also asked about Netflix’s approach to neural MT during the closing panel. The Localization Manager said that currently “it is not part of the strategic plan but we are certainly aware of it.”
Who’s Doing Dubbing?
Smith also explained that the economics of dubbing vs subbing are very different, but highlighted the fact that Netflix sees real value in providing customers with choice. “Choice is more important than a particular preference”, Smith said, and the Netflix localization model is not based on the traditional idea of particular countries preferring dubbing vs subbing. It’s more personal than that for Netflix and, most notably, it’s not possible to get the real data on preference if there’s not the choice, Smith explained.
Netflix has also recently released a list of their dubbing partners, who are arranged into gold, silver and bronze tiers. A list of these partners is included in the table below:
Download the Slator Media Localization Report 2018 for more actionable insights into the media localization industry.
