Supported Languages

Current coverage

A note on voice selection

The SAUTI API uses voice_id in the URL path to select the target voice and language (e.g. /v1/text-to-speech/sauti-swahili-v1). There is no language field in the request body. See the Voices API to list available voices and their language codes.

Why these four languages

African languages are spoken by over a billion people yet remain significantly underrepresented in mainstream AI speech research. The four languages in our roadmap were chosen by three criteria:

• Speaker population — each is among the most widely spoken languages on the continent, maximising reach per model trained.
• Open data availability — datasets like Google WAXAL, Mozilla Common Voice, and existing academic corpora provide a foundation to build on. We augment these with internally sourced recordings where gaps exist.
• Commercial demand — each language corresponds to major economic regions (East Africa, West Africa, the Horn of Africa) with active fintech, telco, and government deployments that need voice AI.

Roadmap timeline

Kiswahili TTS is live. Kiswahili ASR is planned. Hausa, Yoruba, and Amharic are planned for 2026–2027. Each language will follow the same research-to-production pipeline: dataset curation, fine-tuning, evaluation with native speakers, and staged API rollout.

If your use case requires a language not listed here, email hello@finiflowlabs.com. We prioritise languages based on partner demand.

Kiswahili dialect notes

The current model is trained on Kiswahili Sanifu — the standardised written form used across East Africa. Regional dialects (Coastal Swahili, Congolese Swahili) may show reduced naturalness. We document all known dialect gaps in our open research repository.