Business & Technology

Microsoft unveils three real-time audio and speech models for AI agent development

Key points • Microsoft introduced three audio and speech models for developing interactive voice agents. • The transcription model MAI-Transcribe-2-Streaming delivers first hypotheses in just over 100 milliseconds. • The speech models MAI-Voice-2.1 enable consistent vocal identity with a local accent in 23 languages. • The models are now available to developers at a price starting from $0.54 per audio hour and $15 per million characters. Microsoft (Microsoft) is expanding the …