Brand Voice Lite is available to companies on the Yandex Cloud platform. According to the press service, this is the first such service in Russia.
To synthesize a voice, you need to voice the text in the service interface or upload an already prepared recording to it. The speakers can be, for example, invited actors or business owners and representatives. They can synthesize a voice in different roles: for each additional emotion (for example, sad or aggressive speech), you need to record another 20 minutes of speech. The synthesis model will become available within seven days after uploading the recordings to the service.
Yandex emphasized:
The Brand Voice Lite technology complies with Yandex's principles when working with speech synthesis. Businesses must obtain a person's consent to use a recording of their voice. Only the customer will have access to the synthesis results. At the same time, they agree to the rules for using the service, which limit the scenarios for using the synthesis.
Brand Voice Lite is based on Yandex SpeechKit's own ML models, which are fine-tuned for each client on their datasets. The training data is stored in the company's data centers.