app.py

import gradio as gr
from rvc_infer import download_online_model
import os
import re
import random
from scipy.io.wavfile import write
from scipy.io.wavfile import read
import numpy as np
import yt_dlp
import subprocess




def download_model(url, dir_name):
    output_models = download_online_model(url, dir_name)
    return dir_name


uvr_models = {
        'BS-Roformer-Viperx-1297.ckpt': 'model_bs_roformer_ep_317_sdr_12.9755.ckpt',
        'MDX23C-8KFFT-InstVoc_HQ.ckpt': 'MDX23C-8KFFT-InstVoc_HQ.ckpt',
        'BS-Roformer-Viperx-1053.ckpt': 'model_bs_roformer_ep_937_sdr_10.5309.ckpt',
        'Mel-Roformer-Viperx-1143.ckpt': 'model_mel_band_roformer_ep_3005_sdr_11.4360.ckpt',
        'Kim_Vocal_2.onnx': 'Kim_Vocal_2.onnx',
        'UVR-De-Echo-Aggressive.pth': 'UVR-De-Echo-Aggressive.pth',
}





output_format = [
    'wav',
    'flac',
    'mp3',
]

mdxnet_overlap_values = [
    '0.25',
    '0.5',
    '0.75',
    '0.99',
]

vrarch_window_size_values = [
    '320',
    '512',
    '1024',
]






def download_audio(url):
    ydl_opts = {
        'format': 'bestaudio/best',
        'outtmpl': 'ytdl/%(title)s.%(ext)s',
        'postprocessors': [{
            'key': 'FFmpegExtractAudio',
            'preferredcodec': 'wav',
            'preferredquality': '192',
        }],
    }

    with yt_dlp.YoutubeDL(ydl_opts) as ydl:
        info_dict = ydl.extract_info(url, download=True)
        file_path = ydl.prepare_filename(info_dict).rsplit('.', 1)[0] + '.wav'
        sample_rate, audio_data = read(file_path)
        audio_array = np.asarray(audio_data, dtype=np.int16)

        return sample_rate, audio_array

def roformer_separator(roformer_audio, roformer_model, roformer_output_format, roformer_overlap, roformer_segment_size, mdx23c_denoise, mdxnet_denoise, vrarch_tta, vrarch_high_end_process):
  files_list = []
  files_list.clear()
  directory = "./outputs"
  random_id = str(random.randint(10000, 99999))
  pattern = f"{random_id}"
  os.makedirs("outputs", exist_ok=True)
  write(f'{random_id}.wav', roformer_audio[0], roformer_audio[1])
  full_roformer_model = roformer_models[roformer_model]
  prompt = f"audio-separator {random_id}.wav --model_filename {full_roformer_model} --output_dir=./outputs --output_format={roformer_output_format} --normalization=0.9 --mdxc_overlap={roformer_overlap} --mdxc_segment_size={roformer_segment_size}"
    
    if mdx23c_denoise:
        prompt += " --mdx_enable_denoise"
  
    if mdxnet_denoise:
        prompt += " --mdx_enable_denoise"


    if vrarch_tta:
        prompt += " --vr_enable_tta"
    if vrarch_high_end_process:
        prompt += " --vr_high_end_process"

  os.system(prompt)

  for file in os.listdir(directory):
    if re.search(pattern, file):
      files_list.append(os.path.join(directory, file))

  stem1_file = files_list[0]
  stem2_file = files_list[1]

  return stem1_file, stem2_file





CSS = """
"""

with gr.Blocks(theme="Hev832/Applio", fill_width=True, css=CSS) as demo:

    with gr.Tabs():
        with gr.Tab("inferenece"):
            gr.Markdown("in progress")
        with gr.Tab("Download model"):
            gr.Markdown("## Download Model for infernece")
            url_input = gr.Textbox(label="Model URL", placeholder="Enter the URL of the model")
            dir_name_input = gr.Textbox(label="Directory Name", placeholder="Enter the directory name")
         
            download_button = gr.Button("Download Model")
        download_button.click(download_model, inputs=[url_input, dir_name_input], outputs=url_input)

        with gr.Tab("UVR5"):
            roformer_model = gr.Dropdown(
                label = "Select the Model",
                choices=list(uvr_models.keys()),
                interactive = True
            )
            roformer_output_format = gr.Dropdown(
                label = "Select the Output Format",
                choices = output_format,
                interactive = True
            )
            roformer_overlap = gr.Slider(
                minimum = 2,
                maximum = 4,
                step = 1,
                label = "Overlap",
                info = "Amount of overlap between prediction windows.",
                value = 4,
                interactive = True
            )
            roformer_segment_size = gr.Slider(
                minimum = 32,
                maximum = 4000,
                step = 32,
                label = "Segment Size",
                info = "Larger consumes more resources, but may give better results.",
                value = 256,
                interactive = True
            )
            mdx23c_denoise = gr.Checkbox(
                label = "Denoise",
                info = "Enable denoising during separation.",
                value = False,
                interactive = True
            )
        with gr.Tab(" Credits"):
            gr.Markdown(
                """
                this project made by [Blane187](https://huggingface.co/Blane187) with Improvements by [John6666](https://huggingfce.co/John6666)
                """)











demo.launch(debug=True,show_api=False)