Open Oliver-Pikett opened 2 months ago
A better approach would be to return a struct (I was going to raise a ticket for this) - something like:
(pseudo code - won't compile, but a good starter)
struct VADResult {
start_period: 0
end_period: 1000
confidence: 0.7856
voice_detect: true
}
so usage would look something like....
let my_vad = Silero::init( /* whatever params to initialise */);
let result = my_vad.detect(&some_array_with_1_second_of_audio);
println!("confidence: {}, detect_result: {}", resul.confidence, result.voice_detect);
I think that's on its way to be coming a very usable API
make an option for whether the user wants the output to be in the form of booleans or probabilities and for the voice activation threshold when making a session.