On Raspberry PI 3B+, we need to wait a few time between the keyword detection and start the speech. This is annoying because we don't know when to talk.
If we start too soon, the beginning of the speech is cropped.
If we start too late, the speech detection ends without having heard anything.
On Windows it doesn't happen because my machine is probably more performant and this time is reduced to 0.
On Raspberry PI 3B+, we need to wait a few time between the keyword detection and start the speech. This is annoying because we don't know when to talk.
If we start too soon, the beginning of the speech is cropped. If we start too late, the speech detection ends without having heard anything.
On Windows it doesn't happen because my machine is probably more performant and this time is reduced to 0.