app.py

import gradio as gr
import torch
import numpy as np
from torchaudio.functional import resample

from processAudio import upscaleAudio

class Object(object):
    pass

with gr.Blocks(theme=gr.themes.Default().set(body_background_fill="#CCEEFF")) as layout:
    with gr.Row():
        gr.Markdown("<h2>Broadcast Audio Upscaler</h2>")
    with gr.Row():
        with open("html/directions.html", "r") as directionsHtml:
            gr.Markdown(directionsHtml.read())
    with gr.Row():
        modelSelect = gr.Dropdown(
            [
                ["FM Radio Super Resolution","FM_Radio_SR.th"],
                ["AM Radio Super Resolution (Beta)","AM_Radio_SR.th"] 
            ],
            label="Select Model:",
            value="FM_Radio_SR.th",
        )
    with gr.Row():
        with gr.Column():
            audioFileSelect = gr.Audio(label="Audio File (Mono or Stereo, Max 6 Minutes):",sources="upload", max_length=360)
        with gr.Column():
            audioOutput = gr.Audio(show_download_button=True, label="Restored Audio:", sources=[], max_length=360)
    with gr.Row():
        with gr.Column():
            submit = gr.Button("Process Audio", variant="primary", interactive=False)
    with gr.Row():
        with gr.Accordion("More Information:", open=False):
            with open("html/information.html", "r") as informationHtml:
                gr.Markdown(informationHtml.read())    

    @audioFileSelect.input(inputs=audioFileSelect, outputs=[submit, audioFileSelect])
    def audioFileSelectChanged(audioData: gr.Audio):
        #Audio exists and is mono or stereo
        if audioData is None:
            return gr.update(interactive=False), None
        if len(audioData[1].shape) == 1:
            return gr.update(interactive=True), audioData
        if audioData[1].shape[1] > 2:
            gr.Warning("Audio with more than 2 channels is not supported.")
            return gr.update(interactive=False), None
        return gr.update(interactive=True), audioData
        

    @submit.click(inputs=[modelSelect, audioFileSelect], outputs=audioOutput)
    def processAudio(model: gr.Dropdown, audioData: gr.Audio):
        if audioData is None:
            raise gr.Error("Load an audio file.")
            return None
        elif len(audioData[1].shape) == 1: #Convert mono to stereo
            lrAudio = torch.tensor(np.array([
                audioData[1].copy().astype(np.float32)/32768, 
                audioData[1].copy().astype(np.float32)/32768
            ]))
        elif audioData[1].shape[1] > 2:
            raise gr.Error("Audio with more than 2 channels is not supported.")
            return None
        else: #re-order channel data from [samples, 2] to [2, samples]
            lrAudio = torch.tensor(audioData[1].copy().astype(np.float32)/32768).transpose(0,1)
        if audioData[0] != 44100:
            lrAudio = resample(lrAudio, audioData[0], 44100)
        hrAudio=upscaleAudio(lrAudio, "models/" + model)
        hrAudio=hrAudio / max(hrAudio.abs().max().item(), 1)
        outAudio=(hrAudio*32767).numpy().astype(np.int16).transpose(1,0)
        return tuple([44100, outAudio])

layout.launch()