app.py

import gradio as gr
import numpy as np
from vad_utils import get_speech_probs, make_visualization, probs2speech_timestamps


def process_audio(audio_input, model):
    wav = np.array(audio_input)
    probs = get_speech_probs(wav, model, sampling_rate=16_000)
    return make_visualization(probs, 512 / 16_000)

def process_parameters(probs, threshold, min_speech_duration_ms, min_silence_duration_ms, window_size_samples, speech_pad_ms):
    return probs2speech_timestamps(probs, threshold, min_speech_duration_ms, min_silence_duration_ms, window_size_samples, speech_pad_ms)

def main():
    model = None #load_your_model()  # replace with your model loading code

    with gr.Blocks() as demo:
        with gr.Row():
            audio_input = gr.inputs.Audio(type="filepath")
            button1 = gr.Button("Process Audio")
            figure = gr.outputs.Image()

        button1.click(process_audio, inputs=[audio_input, model], outputs=figure)

        with gr.Row():
            probs = gr.State(None)
            threshold = gr.inputs.Number(label="Threshold", default=0.5, minimum=0.0, maximum=1.0)
            min_speech_duration_ms = gr.inputs.Number(label="Min Speech Duration (ms)", default=250)
            min_silence_duration_ms = gr.inputs.Number(label="Min Silence Duration (ms)", default=100)
            window_size_samples = gr.inputs.Dropdown(label="Window Size Samples", choices=[512, 1024, 1536], default=1536)
            speech_pad_ms = gr.inputs.Number(label="Speech Pad (ms)", default=30)
            button2 = gr.Button("Process Parameters")
            output_text = gr.outputs.Textbox()

        button2.click(process_parameters, inputs=[probs, threshold, min_speech_duration_ms, min_silence_duration_ms, window_size_samples, speech_pad_ms], outputs=output_text)

    demo.launch()

if __name__ == "__main__":
    main()