How to generate speech
1
Describe what you need
Tell the agent what text to convert. You can type the text directly or ask the agent to write it first.
2
Choose a voice (optional)
If you do not specify a voice, the agent uses Rachel (the default). To use a different voice, mention it in your prompt.
You can also use a cloned voice by its ID. See Voice Cloning.
3
Review and download
The audio file appears in the chat with a waveform player. Click play to preview, then click Download to save the MP3 to your device.
Example prompts
Phone greeting for a small accounting firm
Voiceover for a bakery's Instagram Reel
Explainer narration for a fitness studio's website
What you cannot do
- You cannot control speed, pitch, or emphasis. The voice model determines pacing naturally.
- You cannot add pauses or breaks at specific positions in the text.
- You cannot output in formats other than MP3 or WAV.
- You cannot mix speech with background music in one step. Generate them separately.
Next steps
Multi-Speaker Dialogue
Generate conversations between multiple speakers.
Voice Cloning
Clone a voice from audio samples to use for speech generation.