Arguments

Next to mymodel.pb a file mymodel.pb.args containing the model parameters is required. Those are defined by the model itself and unless you are making your own model there's no need for them to be changed as a lot of these refere to very internal precise settings or even need a new model to work.

parameter: type -> default. Description.

buffer_t: f32 -> 1.5. Input size of audio. Wakeword must fit within this time.

window_t: f32 -> 0.1. Time of the window used to calculate a single spectrogram frame.

hop_t: f32 -> 0.05. Time the window advances forward to calculate the next spectrogram frame.

sample_rate: u16 -> 16000. Input audio sample rate.

sample_depth: u8 -> 2. Bytes per input audio sample, two for 16 bit.

n_fft: u16 -> 512. Size of FFT to generate from audio frame.

n_filt: u8 -> 20. Number of filters to compress FFT to.

n_mfcc: u8 -> 13. Number of MFCC coefficients to use.

use_delta: bool -> false. If true, generates "delta vectors" before sending to network.

threshold_config: Vec<(f32, f32)> -> (6.0,4.0). Output distribution configuration automatically generated from precise-calc-threshold.

threshold_center: f32 -> 0.2. Output distribution center automatically generated from precise-calc-threshold.

vectorizer is ignored.