Generate Speech turns a written script into spoken audio in the voice you pick, so you can add narration or dialogue without recording anything yourself.
Generating speech is available in Tool mode using any model that supports text to speech.
- Select Tool in the left-hand sidebar
- At the top of the left panel, select Audio
- Open the model picker next to the Generate button at the bottom of the panel
- Select a model with Text to Speech next to its name
- Type or paste your script into the text field at the top of the panel
- Below your script, select a voice. You can select the play button to next the voice name to preview it.
- Select Generate.
The audio clip will in your session, to the right of the panel. Select the play button on the clip to listen to it.
FAQ
How do I control the tone, emotion or pacing of a line?
Start with the voice. Each stock voice is described by traits like calm, intense or hyped, so choose one that already matches the mood you want. From there it depends on the model. Eleven v3 reads emotion and pacing from the script itself, and you can add inline tags such as [whispers], [laughs] or [shouts] wherever you want the delivery to change. Seed Audio 1.0 supports performance tags like [excited] or [sarcastic], lets you write multi-speaker scripts by labeling each speaker, can match a voice's timbre or accent from up to 3 audio references, and has speech rate, loudness and pitch sliders in its settings panel.
Can I use a custom voice?
Yes, on Pro plans and higher, with models that support custom voices. Once a voice is trained, select the Custom tab in the voice list to find it, then select it like any other voice. See How to create a custom voice for requirements and steps.
Can I generate speech in Agent or Workflows?
Yes. This guide covers generating speech in Tool mode, but you can also collaborate with Agent to generate speech and voiceovers, or generate speech using audio models inside a Workflow. See Creating with Runway Agent and Introduction to Workflows to get started with each.