Sahara v2.5 brings code-switching to Africa
Intron’s Sahara v2.5 adds bilingual speech recognition across 12 African language pairs, a Kinyarwanda-English-French trilingual model, and code-switched text-to-speech across 13 languages. It expands Sahara to 63 total supported languages.
Sahara v2.5 targets a major blind spot in global voice AI: people rarely speak one language perfectly from start to finish.
- –Code-switching support could improve voice banking, healthcare, legal transcription, and customer-service tools across Africa
- –The trilingual Kinyarwanda-English-French model is especially relevant for Rwanda’s multilingual workflows
- –Intron reports lower word-error rates than several competing systems on its preliminary code-switched speech benchmarks
- –Developers should independently validate performance across accents, noisy audio, dialects, and production latency before relying on the model for high-stakes applications
DISCOVERED
1h ago
2026-08-26
PUBLISHED
2h ago
2026-08-26
RELEVANCE
AUTHOR
IntronHealth