VK Video has implemented new artificial intelligence algorithms to improve speech recognition and automatic subtitle generation. Speech recognition accuracy has increased by 25%, and neural networks now recognize thousands of new words, including popular memes, proper names, acronyms, and specialized terms.
Automatic subtitles are created using machine learning, which allows not only converting speech to text, but also correctly placing punctuation marks and synchronizing text with video. The system goes through several stages of processing: it removes extraneous noise, converts speech to text, and then punctuation and denormalization models turn it into an easy-to-read format.
In the near future, AI will also be able to separate the speech of different speakers, which will make subtitles even more convenient to understand. This feature is becoming increasingly popular among VK Video users, especially among people with hearing impairments and those who watch videos in conditions where they cannot turn on the sound.
Over the past month, the proportion of users using subtitles in the web version of VK Video has grown by 28%, and now 11% of the entire platform audience uses this feature.
Read materials on the topic:
Viewer controls the blogger: VK Video was the first in Russia to launch interactive content
VK Video has a new update for Android tablets
VK officially launched the VK Video platform: you can watch videos even without registration