Updates
Progress notes and research updates as ÒreAyò develops. Updated periodically with honest reporting on what's working, what's challenging, and what comes next.
Speech Recognition Moves Into Internal Validation
Context
ÒreAyò is being built to understand spoken Nigerian languages in ordinary household settings. Speech recognition is one of the first foundations for that experience, because the assistant must listen before it can respond usefully.
What We're Working On
The first working speech recognition capability is now being validated internally for Yorùbá, Ìgbò, and Hausa. This means ÒreAyò has moved beyond only preparing the research foundation and into testing whether spoken language can be transcribed reliably enough to support future product work.
Why This Is Hard
Nigerian speech is rich and varied. People speak with different accents, mix languages naturally, record in noisy places, and use expressions that do not fit neatly into clean benchmark assumptions.
A useful assistant must handle that reality carefully, not simply work in ideal conditions.
What We've Learned So Far
Foundations matter
Careful data preparation and validation were necessary before useful speech recognition work could begin.
Assumptions changed
An earlier research path did not meet the reliability needed, so the work was re-scoped rather than forced forward.
Validation must widen
Controlled research results are useful, but they are not the same as everyday speech from real families and communities.
What This Enables Later
Speech recognition is not the whole assistant, but it is a necessary step toward spoken conversations. Later, this work can support asking questions by voice, receiving spoken responses, and building interactions that feel natural in Nigerian languages.
Current Status
Internal validation. The full ÒreAyò assistant is not publicly available yet, and the speech recognition capability still needs real-world evaluation before broader access.
Now true
Controlled speech recognition validation exists for three languages.
Not yet true
ÒreAyò is not a public voice assistant or open demo.
Next focus
Real-world audio, quality improvement, and the next voice capabilities.
What We're Not Sharing
We are not publishing training data, model details, or raw benchmark numbers here. That protects cultural data, avoids misuse, and prevents internal validation from being mistaken for a production claim.
ÒreAyò is being built carefully, with language, trust, and real household use in mind. This is progress, not completion.
Previous context
Earlier work focused on the foundations needed for speech research: preparation, experimentation, and validation discipline. That foundation now supports the first internally validated speech recognition capability.
Want to follow along?
Join the waitlist to receive updates as ÒreAyò progresses through research and development.
Join the waitlist