Now, in just a second, VKontakte social network users can decrypt audio recordings up to two minutes long, which is 99 per cent of all audio sent by users. To tell a long story, one can record a voice for up to an hour, NewsNN reports.
It is specified that VK launched the recognition technology in June 2020. Since that moment, the audience that uses voice has grown by 10 per cent (to 33 million people a month). It is stated that on average, each user of the service reads 8 transcripts per day. It takes about a second to recognize a message.
It is noted that the technology is a proprietary development of the VK Team. The algorithms automatically break long audio into short blocks, decode them all at the same time, and then combine them back into one message; the meaning of what was said is not distorted (for this, the Voice Activity Detection technology is used).