TV Center translated 77 thousand hours of video content from audio to text

2022: Translation from audio to text format of large media archive

On November 24, 2022, the federal TV channel TV Center announced that it had translated a large media archive on the Yandex Cloud cloud platform from audio to text format. This is 50 terabytes or more than 70 thousand hours of TV shows, TV shows and documentaries. By transcribing, the channel set up a quick search of archives and began to use media content more efficiently.

Previously, video archives were processed by employees of the TV channel. They manually marked the content with special search tags, so the completeness and quality of the markup were very low. To process 77 thousand hours of video, they would need at least 13 years of continuous viewing. In the cloud, TV Center not only simplified archive searches for employees, but also reduced the number of incidents related to misuse of content.

To transcribe content, the channel uses the cloud service for speech synthesis and recognition Yandex SpeechKit. The technology allows you to generate tags to search for content by events, locations, names. In one month, it was possible to transcribe the entire archive and configure the automatic processing of new content. In the future, TV Center plans to set up a search for content by season, weather and crew.