pipecat
pipecat copied to clipboard
VADAnalyzer falls behind the realtime audio when the chunk size is larger than the size of a VAD frame
A potential bug here: https://github.com/pipecat-ai/pipecat/blob/e982b3d9196acf61c9d029aad48b7ec2b49531a2/src/pipecat/audio/vad/vad_analyzer.py#L91
VADAnalyzer.analyze_audio is triggered by an input audio chunk. Every time it's triggered, only a VAD frame length of audio is processed. When the input chunk is larger than a VAD frame, the VADAnalyzer will fall behind the realtime audio.
Can you provide more information on this? Including a single file repro case with supporting logs?