WaZoBia puts 15 Nigerian-language speech models in one studio
The new Hausa, Igbo and Yoruba hub combines voice cloning, conversational speech generation and transcription, with documented OpenAI-compatible routes.
Three speech workflows in one place
Samuel Olubukun has deployed WaZoBia Sovereign Audio Studio, a community-built Hugging Face Space bringing Hausa, Igbo and Yoruba speech tools into one interface. The new hub groups 15 model options into three tabs: six XTTS-v2 voice-cloning checkpoints, three Chatterbox conversational text-to-speech models and six Moonshine-Tiny transcription models.
The release’s appeal is the shared workflow. Users can select a language-specific checkpoint, use a supplied platform voice or provide reference audio, and move between synthesis and transcription without visiting a separate demo for each model. The English guide also documents OpenAI-compatible speech and transcription routes, making the suite relevant to developers exploring Nigerian-language audio applications.
A new interface, with access confirmed
Repository history dates the first studio deployment to October 14 at 07:48 UTC, followed by compatibility patches. During editorial review, the browser loaded the studio’s three tabs, while its public health and model-list endpoints returned HTTP 200 and identified 15 supported models. Those checks confirm an accessible interface and registry; AITrending did not run speech generation or measure transcription accuracy, voice quality or full compatibility with OpenAI clients.
This is a new unified studio, not evidence that all 15 underlying checkpoints were first released today. The author presents it for non-commercial research and describes separate terms for its component models and training datasets. Its voice policy requires consent before cloning a living speaker. Builders should review those terms before choosing the hub for a production workflow.