I can help with that 👍 It sounds like the issue isn’t Cartesia itself, but how Vapi is handling the voice settings sometimes the default config or model version makes it sound more robotic, and Vapi doesn’t expose all the fine-tuning controls in the dashboard. You usually don’t need a full custom webhook unless you’re trying to dynamically control voice params (like stability, style, temperature, etc.) during the call that can require passing custom values via API instead of relying on the default setup. Are you selecting the model directly in Vapi, or are you injecting the voice config through the API when creating the call?