AssemblyAI

AssemblyAI text-to-speech uses the same API key as AssemblyAI speech-to-text. In the portal, open the AssemblyAI speech credential and check Use for text-to-speech.

{
  "verb": "say",
  "text": "Hi, thanks for calling. How can I help you today?",
  "synthesizer": {
    "vendor": "assemblyai",
    "language": "english",
    "voice": "aai-jane"
  }
}

Voices and languages

The model is assemblyai-tts-1. Set language to AssemblyAI’s own language name: english, spanish, german, italian, portuguese or french. jambonz passes it to AssemblyAI unchanged, so a code such as en-US is rejected. Leave language out and the voice speaks its own language.

The portal lists the built-in library voices (aai-jane, aai-alba, aai-anna, aai-charles, aai-eve, aai-george, aai-jean, aai-mary, aai-michael, aai-paul, aai-vera). Each one speaks all six languages. You can also use any voice_id from the AssemblyAI voice catalog, or a voice you have cloned, as voice. This works with a slug or a UUID.

Options

Pass these in synthesizer.options, or in the credential’s Extra Options.

inactivity_timeout

number

Streaming only. Sets how many seconds a quiet session waits before AssemblyAI closes it, from 5 to 120. The default is 120. jambonz sends a keep-alive every 60 seconds, so the session stays open between turns.

extend_biometric_consent

boolean

Renews the biometric consent of a cloned voice. Only the voice’s owner can set this.

speed

number

Only say reads this, and AssemblyAI currently accepts only 1. Streaming ignores it.

AssemblyAI does not yet support SSML, pronunciation dictionaries or emotion tags.

Data residency

The credential’s TTS data residency setting picks the region: US (tts.us.assemblyai.com, the default) or EU (tts.eu.assemblyai.com). Over the API, set api_uri on the speech credential.

Streaming

AssemblyAI supports TTS streaming. When the caller interrupts the agent, jambonz uses AssemblyAI’s word timings to keep only the words the caller actually heard.