-
|
I just want to classify an 20ms audio whether there are people talking. |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 2 replies
-
|
Hi, 20ms is a too small of an audio chunk size for our VAD - https://github.com/snakers4/silero-vad#how-vad-works |
Beta Was this translation helpful? Give feedback.
-
|
Or maybe longer audio. |
Beta Was this translation helpful? Give feedback.
-
|
Then just split audio into some manageable chunks (i.e. several minutes), batch them, run though the network and plot the results You will have to fiddle a bit with thresholds to produce a simple reliable 1/0 classifier, but I guess some naive approach like a percentage of audio over certain medium threshold will work You can push this into examples afterwards |
Beta Was this translation helpful? Give feedback.
Hi,
20ms is a too small of an audio chunk size for our VAD - https://github.com/snakers4/silero-vad#how-vad-works
For very small chunks - please use WebRTC VAD