Speechmatics supplies speech APIs for companies building voice-driven products rather than a consumer app. Its core speech-to-text service transcribes 55+ languages and dialects, a low-latency mode is built for real-time voice agents, and English text-to-speech is aimed at the response side of a voice-agent pipeline. The company also lists AI translation across 69 language pairs, custom model and custom voice development for enterprise customers, and on-premises deployment options for organisations that cannot send audio to the cloud.
As with any speech-recognition API, accuracy varies by accent, background noise, audio quality and how close a language is to what the model was primarily trained on, so a team building a serious voice product should test the API against its own real audio before committing rather than trusting the language-coverage headline number alone. Enterprise features such as custom voice development and on-premises deployment are priced individually rather than shown publicly, so budgeting for those requires a direct conversation with sales.
Free gives $100 of usage credit with no card required, for up to 2 concurrent real-time sessions across speech-to-text and English text-to-speech. Pro is billed per unit at $0.129, with a stated 20% discount available, supporting up to 50 concurrent real-time sessions and 10 file jobs a second. Enterprise is volume-priced with no rate limits, custom models, custom voice development, on-premises options and a dedicated customer success manager.
