M&E Journal: Optimise Localisation by Introducing Machine Translation During Production
The media and entertainment localisation community is experiencing growing pains due to explosive market growth for content across every platform and language.
# More than 200 streaming services are available around the world. BlueWeave Consulting reports that the streaming market will grow annually by 21 percent between 2021 and 2028 with an estimated total value of $330 billion by 2030.
# The global TV broadcast and media market is expected to grow from $48.47 billion in 2022 to $72.09 billion in 2026 at a CAGR of 10.4 percent, according to Reportlinker.com’s “Broadcast and Media Technology Global Market Report 2022.”
# The theatrical market has a market size of $95.45 billion in 2022 and a revenue forecast of $169.68 billion in 2030 at a CAGR of 7.2 percent, according to Grandview Research’s “Movies and Entertainment Market Size Report, 2030.”
# U.S. streaming platforms aggressively expanded into international markets to gain more subscribers, resulting in greater demand for original language content within the territories these platforms served. To fill this void, content was sourced through licensing and acquisition deals and localised catalog content.
There are no signs that the demand for content will decrease. It’s in the interest of content owners to make streaming more appealing to customers, which requires continual growth in the amount and variety of content.
Subscribers have an insatiable appetite for consuming content anywhere, anytime, and on any device and require compelling content to stay engaged, as well as an extensive library.
During the pandemic, consumer habits changed due to their exposure to great storytelling in foreign-language content provided by OTT services.
Foreign language content offered new narratives that were captivating and original, resulting in subtitled foreign language content entering the mainstream. This phenomenon proved that audiences read when a story is compelling.
Perhaps the greatest shift happened with younger consumers who spend more time on the internet and are more open to foreign language content. According to GWI’s research, “In the UK and U.S., 76 percent of Gen Z/millennials watch foreign language TV shows or films, compared to 56 percent for Gen X/baby boomers.”
Given the current volume of content being localised for streaming, linear/broadcast, and theatrical markets and the increasing demand for localised content, there are simply not enough qualified linguists to translate all this content. Additionally, content is being localised from more source languages and translated into more languages than ever before.
Being a linguist in media and entertainment requires creativity; storytelling capabilities; a deep understanding of cultural nuance, formality, and genre; and language fluency — all while complying with each content owner or distributor’s unique technical specifications and profiles.
Localisation service providers (LSPs) are currently over capacity and can only rely on the capacity available within the freelance market. Even the freelance market is limited because all the LSPs are using the same people and pre-booking time. More linguists will need to be recruited from other sectors or universities and trained.
It will take time to build up our human resources, especially when the Bureau of Labour Statistics predicts only a 2 percent year-over-year growth in translators entering the workforce.
With a tight labour supply, there are only three variables that can adjust to meet the current demand for localised content — time, money, and technology.
# Unfortunately, LSPs don’t have the luxury of more time due to compressed timelines resulting from shrinking release windows and last-minute release changes.
# Spending more money isn’t the answer. With hundreds of billions of dollars being spent on creating content and increased operational costs during a recession, content owners and distributors are more cautious of the bottom line.
# AI-powered machine translation technology is the only effective option to help companies scale to handle unlimited market demand. By focusing linguists’ valuable creative and technical skills on refinement tasks through Machine Translation Post Editing (MTPE), repetitive and tedious tasks can be done through automation and machine translation.
Not only does this allow linguists to focus on what they do best, but it also provides more time devoted to translation accuracy while increasing the throughput of each linguist.
Without AI-powered machine translation, the current state of localisation is unsustainable. We run the risk of linguists leaving the industry due to burnout. We also risk not having enough localised content to meet demand.
Therefore, if we can start the localisation process within the production and post-production window, it will provide additional time to complete localisation with better accuracy, economies of scale, and creative outcomes.
Starting the localisation process during production is crucial because the creatives are still attached to the project, including directors, editors, and most importantly – the actors. This provides a better environment for linguistic directors to collaborate with all parties and ensure that the creative intent is carried through the subtitling, captioning, and dubbing process.
The following are some additional examples where AI-powered machine translation creates efficiencies that support the creative process:
# Speech-to-Text capabilities can support continuity during filming by capturing dialog that was “off script.”
This helps editors and directors choose scenes for the rough assembly and alerts script supervisors/screenwriters where adjustments need to be made to support dialog for scenes shot out of sequence.
# Speech-to-Text capabilities also create efficiencies for non-scripted entertainment where stories are crafted after the footage is captured. Human transcription is a laborious process that speech recognition technologies produce in minutes, helping producers craft storylines quickly before viewing the footage.
# Machine translation can be used by screenwriters when writing jokes or punchlines. A quick machine translation job can indicate how the meaning may be perceived by their wider, international audience.
# For pre-recorded broadcast television content where 20-26 episodes are shot within a nine-month period, machine translation paired with post-editing provides the level of scale necessary to offer the series day- and-date for a global release.
# Synthetic dubbing is a cost-effective alternative that may be appropriate for some content. When actors are still involved with the project, it is much easier to conduct the proper voice sampling necessary to get a good outcome and negotiate their contractual requirements.
The main objections to starting localisation earlier in production lie in two areas. First, the dialogue or scenes may change after its first release (e.g., new theatrical release, streaming release of a feature film).
The beauty of AI-based technologies is that engines can detect these changes. LSPs who worked on the original version can focus on the changed elements.
The second and most difficult issue to address is the selling of rights. While a great deal of planning goes into determining where the content owner will own and sell rights, rights are often negotiated at the last minute and translations may need to be rushed.
Rights may transfer to a distributor, and the localisation elements may not transfer in the sale because they were never created or cannot be found due to the age of the content.
This results in unavoidable, redundant localisation activities but machine translation can also support this challenge.
Given the localisation community’s current challenges, the most cost-effective way to keep up with increased globalisation is by implementing the latest technological advancements.
Statistics show that localisation service providers using machine translation technologies increase their throughput by 30 percent or more and they achieve high- er profitability through cost and time reduction.
Using machine translation earlier in the content lifecycle would provide further efficiencies and could improve the creative process when used strategically.
Before implementing AI-powered machine translation technologies, ensure that you are using a high-quality engine trained with media content that is 100 percent hand-curated by professional translators.
This will produce better quality results than engines trained with imperfect data scraped from the web. Context awareness greatly impacts translation quality by using the information in surrounding sentences to inform the translation vs. translating word for word.
By using specialised engines with these capabilities, your high-quality results will require less post-editing and time – ultimately improving throughput and scale
* By Janice Pearson, Senior Vice President, Sales, Strategy, XL8 *