Speechmatics

Speechmatics provides speech APIs for building voice AI. Its core product is speech to text for real conversations, handling accents, multiple speakers and multilingual speech, in both prerecorded and real time modes. The site lists a general purpose model that transcribes 56 languages without choosing one first and can switch language mid sentence, an Enhanced model aimed at highest accuracy for a single language, a Standard model for cost sensitive volume, a medical model for clinical vocabulary, and a real time model with built in turn detection for voice agents. A text to speech API is also offered. It is aimed at developers and enterprises building captioning, contact center analytics, transcription inside editing software and conversational agents. Speaker diarization, custom dictionary, language hints and smart formatting are available, and the vendor says it can run on device, on premises or in the cloud, does not log customer data for model training by default, and lets customers pick a data processing location in the EU, US or Australia. Compliance listed includes SOC 2 Type II, ISO/IEC 27001:2022, GDPR and HIPAA. Pricing is per audio hour for prerecorded transcription: $0.12 for the general purpose model, $0.38 for Enhanced, $0.23 for Standard and $0.15 for the medical model. Add ons are priced separately, for example translation at $0.49 per hour and summaries at $0.09 per hour. Text to speech is $0.011 per 1,000 characters. New accounts start free with $100 in credits and no card, volume discounts of up to 25 percent apply, and enterprise terms are quoted through sales.