New Generation of Speech Synthesis Technology Unveiled in Russia
Russian company Neuro.net has developed a new generation of speech synthesis technology that uses large language models and diffusion models together to generate speech while accounting for its acoustic characteristics and context.
“New Neuro TTS is built on a modern architecture. Businesses can define how their AI agents sound and integrate speech synthesis into their services through an API,” the company’s press service said, adding that the technology is designed for use in voice-based AI solutions.
The platform is already available in production services. In streaming mode, Neuro TTS starts transmitting audio data 120 milliseconds after a request is made, allowing playback to begin before the entire utterance has been generated.
The technology builds on the company’s experience in contact-center automation, where speech synthesis is part of a full-fledged dialogue with customers.
“I don’t believe that AI development in Russia should be the domain of just a few corporations. Independent technology companies should also have the ambition to set the pace of the market. At Neuro.net, we have been developing our own conversational AI technologies for years. For us, this is both about technological independence and the ability to determine for ourselves what our products can do. They should be able to stand comparison with the best solutions available worldwide,” said Alexander Tsepelev, head of the group of companies that includes Neuro.net.
