Opinion
Your next audience speaks another language
"As AI lowers the cost of translation and dubbing," writes Verbit CEO Yair Amsterdam, "an established channel can enter a new market with years of programming already paid for."
A successful Portuguese-language cooking channel does not need to invent a new format to compete for English-speaking viewers. It already has a host, a recognizable style and a library of recipes that people enjoy watching. Giving that content an English voice is a much smaller undertaking than building a new channel from scratch. AI-assisted translation and dubbing are making that move affordable for more producers.
For years, localization costs have kept many programs out of overseas markets. The prospect of a large audience could justify the investment; a niche series often could not. As adaptation becomes cheaper, more of a catalog becomes worth taking abroad.
I saw a similar shift in our captioning business at Verbit, where lower costs brought new customers into the market. For live sports, we use models trained on sports commentary, with language specialists preparing names and terminology before each game. Automating live captioning allowed us to offer a subscription covering a team's games instead of charging for each hour of human labor. The result was more than savings for existing customers: university sports teams that could never afford human captioners could now make their live broadcasts accessible to more fans.
This changes how content owners weigh funding another production against finding another audience for work already completed. The cost of filming and editing has already been incurred; the new investment is in adapting, distributing and promoting the series abroad. A smaller overseas audience can then be worth pursuing even if it would never justify the cost of a new production.
YouTube's experience gives a sense of the demand. Last year, it reported that creators uploading multilingual audio tracks received, on average, more than a quarter of their watch time in languages other than the video's primary language. There is no need to adapt an entire library before finding out whether people want to watch it. Releasing a few localized episodes can show whether viewers stay and return for more, giving producers a basis for deciding which titles and languages deserve a larger investment.
Local producers, meanwhile, face a different kind of competitor. A channel that is unknown in their market may have spent years building a following elsewhere, with a back catalog and a format refined by audience feedback. Local knowledge still matters, but an established foreign producer can spread its production costs across several markets.
Lower costs matter only if viewers keep watching. A speaker's excitement, hesitation or humor carries meaning as surely as the words do. If the voice keeps talking after the person's lips have stopped moving, the mismatch is hard to ignore. Even when every sentence is accurate, a flat delivery can drain the energy from a good performance.
In our localization work, we partner with dubbing specialists and choose the mix of AI, language experts and voice talent to suit the program. A short instructional video and a drama built around an actor's performance call for different levels of human involvement, from refining an automated voice track to combining AI voices with human performances.
A weak adaptation can also lead to a bad investment decision. If viewers leave because the delivery feels unnatural, a producer may conclude that the show has no audience in that market. A useful test has to preserve what made the original worth watching. Otherwise, it tests the audience's tolerance for a poor dub.
Media companies should revisit the libraries they once judged too expensive to take abroad. Their next big opportunity may already be in the catalog, waiting for an audience halfway around the world. AI-assisted translation and dubbing are making it possible to reach those viewers with work that once seemed destined for a single market.
Yair Amsterdam is CEO of Verbit.
