colossal-glass-16853
12/26/2025, 9:18 AMstt=speechmatics.STT(
api_key=os.getenv("SPEECHMATICS_API_KEY"),
language="tr",
additional_vocab=additional_vocab,
end_of_utterance_silence_trigger=0.2,
operating_point=OperatingPoint.ENHANCED,
max_delay=0.8,
end_of_utterance_mode=EndOfUtteranceMode.FIXED,
transcription_config=TranscriptionConfig(audio_filtering_config={"volume_threshold":3.4, "streaming_mode":True})
),
is volume threshold enough for reducing background noisestall-airline-16756
12/26/2025, 9:18 AMtall-airline-16756
12/26/2025, 9:18 AMvolume_threshold=3.4 in your Speechmatics audio_filtering_config is enough to reduce background noise.
Short answer: I cannot confirm that 3.4 is sufficient. The Speechmatics volume threshold behavior and effective values are not documented here, and LiveKit's built-in noise cancellation (Enhanced Noise Cancellation / BVC) is separate from Speechmatics' internal filtering.
What you can do next:
• Test empirically with sample recordings in your target environment and try several threshold values (lower and higher than 3.4) to see how transcription quality and false triggers change.
• Combine approaches: enable LiveKit's Enhanced Noise Cancellation for the transport layer (WebRTC/SIP) to reduce background noise before it reaches Speechmatics.
• If you want, tell me: 1) a short clip description (type of noise), and 2) whether you can enable LiveKit noise cancellation — I can suggest test values and a simple test plan to find an optimal threshold. Sources: Using Noise Cancellation for SIP and WebRTC Participants | Speechmatics STT plugin guide | LiveKit docs