mod_realtime_ai handles realtime audio, provider connectivity and session lifecycle
required to bring speech-to-speech AI directly into FreeSWITCH calls.
mod_realtime_ai is built as a native FreeSWITCH module. It captures call audio, connects the session to the selected realtime AI provider and plays generated speech back into the call without requiring a separate audio bridge service.
uuid_realtime_aiUse OpenAI, Google Gemini, ElevenLabs or Azure Voice Live through the same FreeSWITCH module. Provider-specific protocol handling stays inside mod_realtime_ai, so your call logic does not need to be redesigned when you select a different provider.
Caller audio is streamed to the AI provider while generated audio is returned and played into the FreeSWITCH call. Both directions are handled as part of the same realtime session for natural conversational voice applications.
Telephony audio and realtime AI services do not always operate at the same sample rate. mod_realtime_ai automatically converts audio between the FreeSWITCH channel rate and the rate required by the selected provider.
<profile name="sales">
<param name="provider" value="openai"/>
<param name="config" value="sales.json"/>
</profile>Define reusable profiles for different providers, applications and AI configurations. FreeSWITCH only needs the profile name when the realtime session starts, keeping provider configuration separate from your call logic.
The module manages the realtime session from connection through audio processing and cleanup. Transport, playback and session state are coordinated inside the module so applications can focus on call flow and AI behavior.
Start with the free version, evaluate without channel limits for 30 days, or explore the documentation first.