Skip to main content
Voice generation lets an Agent turn text into spoken audio when the capability is available in the current workspace. Give it the exact script, intended audience, language, pronunciation guidance, and tone. The generated audio is an output artifact and should be listened to before distribution. In the Agent builder, enable the voice or audio generation capability shown under Tools. Ask for one short sample first. Review names, numbers, pronunciation, pauses, and whether the delivery matches the requested style. If a voice profile or reference recording is used, ensure it is authorized and appropriate for the intended audience. Generation quality and supported formats depend on the active model and workspace. Do not assume that a transcript, recording, or generated voice has been approved for publication. Save or download the output only after checking it end to end.

​
Speech Generator

The Speech Generator can produce single-speaker narration or multi-speaker dialogue with configurable voice, language, and speaking style. The source specification lists up to 10,000 characters for single-speaker speech, up to 5,000 dialogue lines, support for 21 languages, and MP3 output. Check the current capability panel for active limits and formats. Enable it in the Agent builder; no separate provider setup is described.