SPEECH

Speech-optimized VAD using Silero neural network.

State-of-the-art accuracy with small model size (~628KB). Recommended for speech detection. Requires ai + ai-models dependencies.