> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agents.labs.bandwidth.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Choose a voice

> Select a speech provider, model, and voice for your agent.

Open **AI connections → Voice** to set the workspace default, or enable a Voice override on an individual agent.

1. Choose a provider and a saved credential.
2. Select a model and voice available to that account.
3. Set the language and any supported voice options.
4. Save and run a browser conversation to check pronunciation, pace, and response time.

If you enter a voice ID manually, copy it from the same provider account used by the credential. A voice ID from another provider will not work.

For a realtime conversation, choose the voice in the realtime connection. Use [custom recordings](/voice-agent/pre-recorded-audio) when a workflow needs prepared audio.