In a React Native application, is it possible to pass audio directly to the avatar instead of text?

In our current implementation, the avatar speaks based on the text input provided. However, we recently modified the flow to improve interactivity. Now, instead of manually entering text, we're using speech-to-text conversion — the user speaks, the input is transcribed, and then the avatar speaks that text.

To make the experience more interactive, we've enabled an "always-on" microphone mode. However, we're encountering an issue: the microphone is picking up the avatar's voice as well, leading to a feedback loop and echo.

Our goal is to send audio directly to the avatar for processing, bypassing text altogether. Additionally, we need echo cancellation similar to what WebRTC offers. Unfortunately, we're not using WebRTC in the current implementation, so we’re stuck on how to effectively achieve this.

Is there a way we can feed raw audio directly to the avatar and implement echo cancellation to avoid this looping issue?