Noise and Voice Breaking in HeyGen LiveAvatar for Long Sentences

Hello HeyGen team,

I’m facing an issue while using the HeyGen LiveAvatar Web SDK.

Issue:

Short sentences → avatar speaks clearly

Long sentences → avatar audio has noise and voice distortion

Important:

The original TTS audio is clear and noise-free

The issue appears only after passing audio to LiveAvatar

Observations:

Issue happens only for long audio

Feels related to audio chunking, buffering, or timing during streaming

Audio is PCM (24kHz, mono, s16le)

Expected behavior:
Long sentences should play smoothly, same as short ones.

Has anyone faced a similar issue or can suggest best practices for handling long audio with repeatAudio()?

Thanks in advance.