VTB has developed a branded voice for a smart financial assistant using technology from Yandex.Cloud - SpeechKit Brand Voice, the bank told. VTB promises that the voice assistant will be able to advise users on financial services in a unique, fully synthesized voice.
Now the voice developed by the bank is used in the voice assistant Alice, within a year it will appear in the VTB Online mobile application. The VTB press release says that users will be able to ask the assistant any questions about products and services - for example, the assistant will tell you about the conditions for issuing accounts and credit cards, tell you about existing bonus options, etc. The voice of a Russian theater, film and dubbing actor was chosen for voice acting Vladimir Zaitsev, who dubbed Robert Downey Jr. and Jason Statham, as well as Geralt of Rivia in the first part of the Witcher series of games.
The company notes that consultation with an assistant through Alice's skill will be confidential. To activate the voice assistant, you need to tell Alice “Start the VTB Assistant skill” through the Yandex application or in Yandex.Station smart speakers.
The assistant's voice was synthesized using Yandex SpeechKit Brand Voice technology. It was developed on the basis of ML speech models - the bank claims that the technology allows you to create full-fledged digital copies of a person's voice. The models were trained using transformers, since they can process the entire array of text at once, unlike sequential neural networks. Due to this, development time is reduced and the quality of synthesis and recognition of human speech is improved, VTB says. Yandex.Cloud specialists trained the ML model on a dataset of dozens of hours of real person voice recordings and handed it over to VTB developers.
The bank noted that the model will be regularly trained, which will improve the quality of interaction with users and add new scenarios.
Yandex SpeechKit Brand Voice was introduced by the company in July 2021. With the technology, third-party companies will be able to synthesize voice for voice assistants, automated call centers, surveys, telemarketing, and quality control.
bbabo.Net