When a user speaks into their microphone, the avatar's pre-configured visemes respond in real-time, creating a convincing illusion of speech. This feature enables the virtual character to mimic the user's speech patterns, providing an immersive and realistic conversational experience. This is particularly impactful in platforms like Hyperfy and VRChat, where the seamless blend of audio input and avatar animation enriches social interactions, making digital communication feel more natural and engaging.

When setting up a vrm youll typically connect a shapekey or blendshape to each vowel. I am not sure on the technical of implementing this but vrms have default support for them.