AI-Powered Language Coach Helps Learners Practice Speaking on Their Own
A team from the startup studio at Novosibirsk State University has developed TalkTuner, an AI-powered language coach designed to help learners overcome the fear of speaking a foreign language. The developers believe the tool could make conversation practice far more accessible by allowing users to improve their speaking skills independently, without relying on teachers or language clubs.
TalkTuner Aims to Make Language Learning More Conversational
The service builds on the professional experience of linguist Katerina Fomel, who has taught languages for more than a decade. She identified two common obstacles for learners: fear of making mistakes and shyness. Existing AI tools do not always solve those problems. Many international services remain difficult for users in Russia to pay for, while standard chatbots often provide only limited feedback on language errors. TalkTuner takes a different approach by creating a low-pressure environment for speaking practice while offering detailed feedback.
The platform delivers personalized speaking exercises powered by artificial intelligence. It is designed for learners who already understand basic grammar and vocabulary but need regular conversation practice to become more fluent. Today, the AI coach operates as a Telegram chatbot. Users record a spoken message in a foreign language and send it as a voice note. The neural network responds with spoken replies on the selected topic. Beyond keeping the conversation going, the AI coach evaluates spoken language, identifies grammatical and vocabulary weaknesses, and explains users' mistakes.
Multiple AI Models Working Together
The project was developed at the Novosibirsk State University startup studio, where it progressed from an initial concept to a market-ready product model. The creators are university students: Katerina Fomel is pursuing a master's degree, while Kseniya Mikhalkina is an undergraduate student. Their platform combines several AI models into a single service. OpenAI's Whisper handles speech recognition, the Deepgram TTS API generates the AI coach's voice, and the DeepSeek language model manages conversations while identifying language errors.
The architecture also allows developers to swap AI models depending on the learning objective. Participation in the Catalyst accelerator helped the team define the key parameters of its minimum viable product and begin development. The chatbot is now being used by testers and paying customers, with its first commercial sales already completed. TalkTuner also won the A:START business accelerator and earned an invitation to the Academpark business incubator.
TalkTuner is part of the Platforma universitetskogo tekhnologicheskogo predprinimatelstva (University Technology Entrepreneurship Platform), which operates under the federal project "Tekhnologii" (Technologies) within the national project "Effektivnaya i konkurentosposobnaya ekonomika" (Efficient and Competitive Economy). Looking ahead, the development team plans to introduce detailed pronunciation analysis, add a library of video lessons, and release both a mobile app and a web version of the service.
A Boost for Educational Software
The student-led project from Novosibirsk State University reflects a broader shift in educational technology. Instead of conventional chatbots and digital textbooks, developers are beginning to deploy AI-powered voice tutors that can recognize speech, sustain natural conversations, and analyze learners' mistakes. For users, these tools could make speaking practice far more accessible because they can study independently without paying for a private instructor.
At a national level, the technology could stimulate growth in Russia's educational software market while encouraging new products that combine linguistics, speech recognition, and generative AI. In the future, TalkTuner could be integrated into university curricula, language schools, and corporate training programs across Russia. Individual learners could also use it to strengthen their language skills before job interviews, international travel, or language proficiency exams. The service could also find export opportunities in CIS countries, Asia, Africa, and the Middle East, where demand for foreign-language learning remains strong, provided the developers adapt it to local regulatory requirements and user needs.
