Speechmatics · mars 2026
4 mises à jour · mars 2026
En mars 2026, Speechmatics a publié 4 mises à jour suivies. En hausse de 33% par rapport à février 2026. Le thème dominant : content pieces (2). À retenir : « Speech-to-text in production: what 36 years of hard lessons taught me » (31 mars 2026).
Parcourir d'autres mois de Speechmatics
Speechmatics founder Dr. Tony Robinson highlights key challenges in Voice AI production, arguing that LLMs—not ASR—are the primary failure point in real-time pipelines. He emphasizes the need for prov…
The U.S. faces a structural court reporter shortage with ~17,700 professionals, retirements outpacing new entrants, and California reporting 1.7M+ proceedings without verbatim records since 2023. Lega…
Speechmatics published a guide clarifying that no single word error rate (WER) threshold applies universally to legal transcription, as court rules and standards emphasize qualitative accuracy over nu…
Speechmatics introduced the first Arabic–English bilingual speech-to-text model, achieving 6.3% WER on mixed-speech benchmarks and 35% fewer errors than Google. The model supports real-time and batch …
Suivez Speechmatics en pilote automatique