ConversationRelay: Deepgram Nova-3 Speech-to-Text is Now Available

ConversationRelay now officially supports Deepgram’s latest speech-to-text model, nova-3. Deepgram nova-3-general will be the default model (e.g. transcriptionProvider and speechModel combination) for all new customers; current customers will not be impacted. In addition to nova-3-general, the nova-3-medical version is also available. Lastly, we’ve updated ConversationRelay’s “hints” parameter to include support for Deepgram’s newer “keyterms” feature for nova-3.

Support for nova-3’s multilingual language option will come in a subsequent planned release.

To learn more about ConversationRelay please check out the docs here. We can’t wait to see what you build!