Silero VAD is a pre-trained speech activity detector created by Silero Team; see Silero Team (2024) for the model and documentation. VAD stands for “voice activity detection”, but we use “speech activity detection” instead, because Silero VAD detects unvoiced parts of speech as well as voiced parts. The Silero VAD model is compiled into Praat (see Acknowledgments).

For how Silero VAD is used in Praat, as well as its algorithm and settings, see speech activity detection with Silero VAD.