Optionalsimilarity_How closely the synthesized speech should adhere to the original voice it was cloned from.
A number from 0 to 1: higher follows the original more strictly.
OptionalstabilityHow stable the voice is, and how much each generation varies.
A number from 0 to 1: lower is more expressive and more variable between generations, higher is flatter and more repeatable.
How closely an ElevenLabs voice should follow the original it was built from.
The
voice_configof ElevenlabsTtsProvider, mirroring ElevenLabs' own voice settings. Both fields are optional; when one is omitted the Agents API applies its own default.