Developers: | Yandex.Cloud |
Date of the premiere of the system: | 2023/03/07 |
Technology: | Call Centers, Voice Technology |
The main articles are:
- Speech Recognition (Technology, Market)
- Speech technology: On the path from recognition to understanding
- Speech synthesis
The Brand Voice Call Center complements the Brand Voice product line, which also features Brand Voice Self Service and Brand Voice Premium. Brand Voice Self Service is a full-text voice synthesis that will suit any communication with customers and voice acting of text content. And Brand Voice Premium allows you to create unique voices in different roles for marketing, PR campaigns and voice assistants.
2023: Service Presentation
On March 7, 2023, the Yandex Cloud cloud platform introduced the Brand Voice Call Center speech synthesis service. With the help of technology, companies will be able to create unique voices for virtual call center operators in almost real time. In this case, the "robot" can be taught, for example, to contact customers by name or to coordinate the addresses and names of goods in the order. This will allow the business to personalize and revive communication in voice channels. Brand Voice Call Center is already available for companies and is available on request.
The algorithm processes one audio pattern and synthesizes hundreds of the same phrases on its basis, but at the same time it can change individual words in them by script. At the same time, the synthesized speech in the Brand Voice Call Center sounds natural and conveys all the details of the speech of a living person from the template: emotions, intonations, volume changes. As templates, companies can use fragments of telephone records of real operators of their call centers. This is the first such service in Russian, released for commercial use.
To teach Brand Voice Call Center, Yandex Cloud specialists used a datacet with thousands of hours of recordings of various announcers in Russian from open access. Such experience allows you to work with almost any voice without prior preparation. To make speech sound more natural, a transformer architecture was used to train models in the service. Unlike other neural networks, transformers allow you to train ML models in parallel on modern video cards (GPUs) and concentrate on important parts of the text, which increases the quality of synthesis.
For March 2023, the service is already used by a medical company and a large telecom operator that uses Brand Voice Call Center for its customers and sees a 20% increase in conversion in voice sales channels using this technology. And according to the PBX company, personalized speech helps to significantly increase customer loyalty to virtual operators.