Deepgram: Nova-3

Deepgram's highest-performing general-purpose speech-to-text model for synchronous and streaming transcription, recommended for meetings, event captioning, multi-speaker audio, multilingual audio, noisy environments, and far-field audio.