For the complete documentation index, see llms.txt. Markdown versions of documentation pages are available by appending .md to page URLs.
Skip to content

Realtime Voice Families

Developer-facing realtime voice capabilities, with links to family selection, service configuration, and app voice session integration.

GPT Realtime Dynamic Behaviors

Grok Voice

  • Run a single-state workflow in a built-in Grok voice on app voice sessions, inbound phone calls, Playground, and test calls by selecting grok_voice on a service in a workspace without regional data residency. See Grok Voice configuration.

GPT-Live 1.1 Multilingual Alpha

  • Opt a GPT-Live service into gpt-live-1.1-leopard-2-alpha, OpenAI's GPT-Live 1.1 multilingual alpha, by naming it in voice_config.gpt_live.model. See GPT-Live configuration.

Voice Family Discovery

  • Read each voice family's availability for the workspace, its models, voices, and settings schema from GET /v1/{workspace_id}/voice-families instead of hard-coding them. See Discover Voice Families.

GPT-Live

  • Run multi-state Context Graphs with realtime speech by selecting gpt_live on a US service without regional data residency. See GPT-Live configuration.

GPT Realtime Multi-State Workflows

  • Run gpt_realtime with 2 to 12 action states by giving each a unique name and valid exits, setting both initial-state fields to the first state, and disabling dynamic behaviors. See GPT Realtime Multi-State Workflows.

October 31, 2026

  • Run Eleven Realtime on inbound phone calls, Playground, and test calls by setting the service's session_provider and voice. See Eleven Realtime configuration.

Last updated

Was this helpful?