The purpose of the tool is to increase the reach of online video content by taking away language barriers. Journalists can use the system to automatically generate multilingual voice-overs without having to go through a lengthy, expensive translation process. “This way, they can suddenly produce 10 to 15 more videos a day,” said Shishkin, who spoke about the technology at a recent news:rewired conference in London.
How automatic translation and voice-over tool works
The video is automatically translated by running the original English-language script (uploaded by the editors) through Google translate.
The automated translation then requires a human eye to correct errors and adjust linguistic sensitivity, which is done by the BBC’s editors.
The editors select their preferred synthetic voice from a drop-down menu and publish it through the system.
Watch video explanation here:
The BBC started the pilot in December 2015 on BBC Japan – and shortly after that on BBC Russian. The technology will be tested and further developed until September this year, with further plans to expand the project to cover Spanish, Portuguese and Korean-speaking regions.
While it is still too early to fully understand the audience response to the synthetic voices, Shishkin said that “the completion rates for the automatically translated Japanese videos are not worse than those created by humans”. It may be because the country is more accustomed to these type of things and because the synthetic voices are getting better, he explained. “Even Japanese editors working on the Japanese website in Tokyo rated the voices 8 out of 10 for quality.”