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

def process_audio(audio_input):
    wav = read_audio(audio_input, sampling_rate=16_000)
    probs = get_speech_probs(wav, 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(): 
    


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

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

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