Why I Started Sauti Labs

Description and use

The author founded Sauti Labs to bridge the gap between research datasets/models and production‑ready speech AI for African languages, after discovering limitations in existing ASR tools like Whisper.

Problem

Speech recognition models like OpenAI’s Whisper perform poorly on African languages and accents, limiting the ability to build reliable voice services for African communities.

Outcome

The author founded Sauti Labs to develop production‑ready speech and language AI for African languages, addressing gaps in accuracy, licensing, and deployment that existing research resources do not fully cover.

Workflow

• 1. Build Scribe AI using Whisper for ASR.
• 2. Evaluate Whisper’s performance on African languages.
• 3. Search for alternative models and datasets.
• 4. Discover community resources (Masakhane, Maseno, Lelapa AI).
• 5. Identify limitations: lack of models for specific use cases, high error rates, restrictive licenses, hardware constraints.
• 6. Decide to create Sauti Labs to bridge research and product gaps for African languages.

Evidence

In July 2026, I was building Scribe AI, an offline system that listens to a church service and puts the scripture being referenced on screen. Everything in it depends on speech recognition, and for that I used OpenAI’s Whisper.

Source

Whisper’s training data is heavily concentrated in English. Swahili appears with roughly five hours of transcribed speech, while Dholuo—the language that first drew me into this work—is absent from its supported-language list.

Source

Masakhane, a grassroots community focused on strengthening natural language processing research in African languages, has spent years building open datasets, models, and benchmarks.

Source

The products that did exist had been built, sensibly, around particular languages and markets. That is how products should start, but it still left the speech I needed uncovered.

Source