Input
- Required Inputs:
output: URL of the audio recording to evaluate
Output
Result: Count of inappropriate interruptions (integer)Reasoning: Brief explanation of detected interruptions
Supported models: Gemini 2.5 Flash (recommended), GPT-4o Audio Preview, Gemini 2.5 Pro. Gemini 2.5 Flash is recommended for best performance and latency.
Interpretation
- Higher counts: More frequent AI interruptions
- Lower counts: Better turn-taking and conversational etiquette