we are trying to develop emergency call center usecase. Here we are trying to capture the background noises if the user unable to speak, and his emotion also like heavy breathing. But Retell AI not able to capture any background noises when we keep the interruption to low also.
please help to understand and how can I archive this here.
Based on the available documentation, Retell AI’s design focuses on removing background noise rather than capturing it, which creates a fundamental challenge for your emergency call center use case.
Retell AI offers two denoising modes(1):
Remove noise (default) - removes background noise with nearly no distortion to the waveform and no impact on speech-to-text accuracy, but cannot remove loud background speech(1)
Remove noise + background speech - more aggressive mode that removes both background noise and background speech, though it may distort the waveform and result in lower speech-to-text accuracy in certain cases(1)
Both modes are designed to filter out background sounds, not preserve them(1).
The interruption sensitivity setting helps reduce false interruptions from background speech and noise by making the agent more resilient to these sounds(1). Setting it lower (like 0.8) makes the agent less responsive to background noises(1), which is the opposite of what you need for detecting distress signals.
Retell AI’s speech processing architecture uses Voice Activity Detection (VAD) and turn-taking models that are optimized to distinguish speech from non-speech elements(2)(3). The system is built to filter out ambient sounds and focus only on clear speech signals(2)(4).
For an emergency use case requiring detection of:
The current Retell AI architecture works against these requirements, as it’s designed to remove exactly the audio information you need to capture.
I don’t have documentation showing a way to disable noise removal or access raw audio streams through Retell AI. You may want to contact the Retell AI team directly to discuss:
Contact options: