You can map a text-to-speech (TTS) engine and voice per flow to use per language.

Depending on when you create or update the flow, the system opens the flow with a TTS engine as follows:

You cannot set a TTS voice per prompt; this option is only available on a per-flow basis for each supported language of the flow. At runtime, Architect uses the TTS engine and voice you configure for the flow.

1. Create a new flow or open an existing one.
2. Under Settings, click Supported Languages.
3. Do one of the following:
   • Add a new supported language.
   • Locate an existing supported language.
4. In the Text To Speech column for the supported language, do the following:
   1. Under Engine, select the TTS engine for the flow:
      • None: Select this option if the language does not have an available TTS engine or voice.
      • Default: To use the language’s default TTS engine configured for the organization, select this option. 
        Note: If the organization’s default TTS engine or voice changes, and you want the existing flows that use this option to use the new default engine and voice, you must republish the flow.
      • Genesys TTS: To use the built-in Genesys TTS engine, select this option.
      • Genesys Generative TTS: To use the generative TTS engine that offers more human-like TTS voices, select this option.
        Supported generative TTS voicesClick to expand

        Voice model	Provider	Call flows	Bot flows (Virtual Agents, AI guides, and agentic virtual agents)	Voice surveys
        Chirp 3: HD voices	Google Cloud	✓	✓	✓
        Eleven Flash v2.5	ElevenLabs	✓	✓	✓
        Gemini 3.1 Flash TTS	Google Cloud	✓	✓	✓
        Generative voices	Amazon Polly	✓	✓	✓

      • Genesys TTS Connector – An integrated third-party TTS engine: To use your preferred third-party TTS engine configured using the Genesys TTS Connector integration.
      • An installed third-party TTS engine: Select this option if an administrator obtained and installed the third-party TTS engine exists from the Genesys AppFoundry.
   2. Under Voice, select the voice you want to read the text-to-speech.
      Notes:

      • You can select any voice that the selected TTS engine supports. The TTS voice does not need to match the flow’s language. For example, you can use a language as your flow’s default language for which Architect supports speech recognition (“en-GB”), and at the same time, use a different TTS voice (“en-IE”) to meet your customers’ language needs.
      • You can only select a generative voice if the selected TTS engine supports it. Currently, Genesys Generative TTS and Amazon Polly (third-party) engines support generative TTS voices such as en-us, polly-Danielle-Generative. The generative TTS voices are unavailable in secure call flows.
      • You can configure a flow with a language that supports runtime data playback and select a different TTS voice for localized speech. For example, you can use English (“en-US”) as the flow language to enable runtime data playback and select English (“en-IE”) voice to provide an Irish accent.
5. Save the flow.