Ability to Play/Inject Audio During SIP Tool Calls #710
Description
When executing tool calls via the OpenAI SIP interface, there currently doesn’t seem to be a way to play or inject audio to indicate that an action is happening. This kind of feedback would be very helpful in user-facing voice applications, where silence can be mistaken for inactivity.
With the Web SDK, it’s possible to detect pauses and insert audio cues manually, but I haven’t found an equivalent approach for SIP. It would be great if SIP provided a built-in method or event hook to trigger audio output during tool execution.
Proposed Enhancement:
- Add a mechanism within SIP to output audio (e.g., tones, brief clips, TTS) while a tool call is being processed.
- Alternatively, expose events that allow developers to inject audio at specific points in the SIP pipeline.
This feature would significantly improve UX for voice-driven applications relying on SIP.