Sonora: Bringing the Belarusian Language Into Voice AI
Sonora is a Belarusian community-driven voice AI initiative creating the first high-quality open studio dataset for Belarusian text-to-speech technologies.
Today, Belarusian is still poorly represented in global voice AI systems. Existing TTS voices often struggle with stress placement, homographs, phonetics, rhythm, and natural intonation. For users, this means Belarusian AI voices often sound robotic, inaccurate, or unnatural.
Sonora addresses this at the foundation level: we are building a professionally recorded, linguistically curated Belarusian speech dataset focused on the hardest parts of the language, including homographs, correct stress patterns, phonetic precision, and natural human speech.
The project is non-profit and community-funded. The first stage budget is €13,000 and covers text preparation, studio recording, voice talent, sound engineering, post-processing, annotation, quality control, and dataset documentation.
What We Offer To Technology Companies
Sonora can provide companies such as OpenAI, Google, ElevenLabs, Speechify, Meta, and other voice AI teams with:
- an open, high-quality Belarusian studio speech dataset
- linguistic expertise on Belarusian pronunciation, stress, homographs, and phonetics
- access to a motivated Belarusian-speaking community for testing and feedback
- collaboration with AI engineers fine-tuning open-source voice models
- cultural and language expertise that is difficult to build internally for a smaller-language market
- a practical path to improving Belarusian support in voice assistants, TTS, accessibility tools, education, media, and AI agents
For large AI companies, Belarusian may be a small commercial market, but supporting it well is a strong signal of linguistic inclusion, cultural responsibility, and technical quality across low-resource languages.
Why This Matters
Voice is becoming one of the main interfaces between people and technology. If Belarusian is absent or poorly implemented in voice AI, it risks being pushed further out of everyday digital life.
High-quality Belarusian TTS would enable:
- accessible tools for blind and dyslexic users
- audiobooks, podcasts, and media production
- educational content for students and teachers
- Belarusian-speaking AI assistants, chatbots, navigation systems, and interfaces
- language access for the Belarusian diaspora and their children
- better foundations for future speech technologies
Sonora's goal is simple: Belarusian should sound natural, correct, and alive in modern technology.
Team And Support Network
Sonora is led by its founder and supported by a volunteer team and partner network, including voice AI engineers fine-tuning open-source models, linguistic and technical contributors, Dev.by media support, and active community fundraising through Gronka.
Some contributors remain non-public for safety reasons.
What We Are Looking For
We are looking for direct contacts and collaboration opportunities with AI and voice technology companies.
We would like to explore how Sonora's dataset, expertise, and community can help improve Belarusian language support in major AI products. In return, companies can help ensure that Belarusian speakers are not left behind in the next generation of voice-first technology.