app.py

import os
import sys
import gradio as gr
import urllib.parse
from audio_separator.separator import Separator

from argparse import ArgumentParser
from assets.presence.discord_presence import RPCManager, track_presence
from variable import *
from code import *

now_dir = os.getcwd()
sys.path.append(now_dir)

device = "cuda" if torch.cuda.is_available() else "cpu"
use_autocast = device == "cuda"

if os.path.isdir("env"):
    if platform.system() == "Windows":
        python_location = ".\\env\\python.exe"
        separator_location = ".\\env\\Scripts\\audio-separator.exe"
    elif platform.system() == "Linux":
        python_location = "env/bin/python"
        separator_location = "env/bin/audio-separator"
else:
    python_location = None
    separator_location = "audio-separator"


def load_config_presence():
    with open(config_file, "r", encoding="utf8") as file:
        config = json.load(file)
        return config["discord_presence"]

def initialize_presence():
    if load_config_presence():
        RPCManager.start_presence()

initialize_presence()


with gr.Blocks(title="🎵 UVR5 UI 🎵") as app:
    gr.Markdown("<h1> 🎵 UVR5 UI 🎵 </h1>")
    gr.Markdown("If you liked this HF Space you can give us a ❤️")
    
    with gr.Tabs():
        with gr.TabItem("BS/Mel Roformer"):
            with gr.Row():
                roformer_model = gr.Dropdown(
                    label="Select the model",
                    choices=list(roformer_models.keys()),
                    value=initial_settings.get("Roformer", {}).get("model", None),
                    interactive=True
                )
                roformer_output_format = gr.Dropdown(
                    label="Select the output format",
                    choices=output_format,
                    value=initial_settings.get("Roformer", {}).get("output_format", None),
                    interactive=True
                )
            
            with gr.Row():
                roformer_audio = gr.Audio(
                    label="Input audio",
                    type="filepath",
                    interactive=True
                )
            
            with gr.Row():
                roformer_button = gr.Button("Separate!", variant="primary")
            
            with gr.Row():
                roformer_stem1 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    label="Stem 1",
                    type="filepath"
                )
                roformer_stem2 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    label="Stem 2",
                    type="filepath"
                )
            
            with gr.Accordion("Separation by link", open=False):
                with gr.Row():
                    roformer_link = gr.Textbox(
                        label="Link",
                        placeholder="Paste the link here",
                        interactive=True
                    )
                with gr.Row():
                    gr.Markdown("You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)")
                with gr.Row():
                    roformer_download_button = gr.Button("Download!", variant="primary")

            roformer_download_button.click(download_audio, [roformer_link], [roformer_audio])

            with gr.Accordion("Batch separation", open=False):
                with gr.Row():
                    roformer_input_path = gr.Textbox(
                        label="Input path",
                        placeholder="Place the input path here",
                        interactive=True
                    )
                    roformer_output_path = gr.Textbox(
                        label="Output path",
                        placeholder="Place the output path here",
                        interactive=True
                    )
                with gr.Row():
                    roformer_bath_button = gr.Button("Separate!", variant="primary")
                with gr.Row():
                    roformer_info = gr.Textbox(
                        label="Output information",
                        interactive=False
                    )

        with gr.TabItem("Roformer Processing"):
            with gr.Row():
                roformer_segment_size = gr.Slider(
                    label="Segment size",
                    info="Larger consumes more resources, but may give better results",
                    minimum=32,
                    maximum=4000,
                    step=32,
                    value=initial_settings.get("Roformer", {}).get("segment_size", 256),
                    interactive=True
                )
                roformer_override_segment_size = gr.Checkbox(
                    label="Override segment size",
                    info="Override model default segment size instead of using the model default value",
                    value=initial_settings.get("Roformer", {}).get("override_segment_size", False),
                    interactive=True
                )
            with gr.Row():
                roformer_overlap = gr.Slider(
                    label="Overlap",
                    info="Amount of overlap between prediction windows",
                    minimum=2,
                    maximum=10,
                    step=1,
                    value=initial_settings.get("Roformer", {}).get("overlap", 8),
                    interactive=True
                )
                roformer_batch_size = gr.Slider(
                    label="Batch size",
                    info="Larger consumes more RAM but may process slightly faster",
                    minimum=1,
                    maximum=16,
                    step=1,
                    value=initial_settings.get("Roformer", {}).get("batch_size", 1),
                    interactive=True
                )
            with gr.Row():
                roformer_normalization_threshold = gr.Slider(
                    label="Normalization threshold",
                    info="The threshold for audio normalization",
                    minimum=0.1,
                    maximum=1,
                    step=0.1,
                    value=initial_settings.get("Roformer", {}).get("normalization_threshold", 0.9),
                    interactive=True
                )
                roformer_amplification_threshold = gr.Slider(
                    label="Amplification threshold",
                    info="The threshold for audio amplification",
                    minimum=0.1,
                    maximum=1,
                    step=0.1,
                    value=initial_settings.get("Roformer", {}).get("amplification_threshold", 0.7),
                    interactive=True
                )
            with gr.Row():
                roformer_single_stem = gr.Textbox(
                    label="Output only single stem",
                    placeholder="Write the stem you want, check the stems of each model on Leaderboard. e.g. Instrumental",
                    value=initial_settings.get("Roformer", {}).get("single_stem", ""),
                    interactive=True
                )

            roformer_bath_button.click(
                roformer_batch, 
                [roformer_input_path, roformer_output_path, roformer_model, roformer_output_format, 
                 roformer_segment_size, roformer_override_segment_size, roformer_overlap, 
                 roformer_batch_size, roformer_normalization_threshold, roformer_amplification_threshold, 
                 roformer_single_stem], 
                [roformer_info]
            )

            roformer_button.click(
                roformer_separator, 
                [roformer_audio, roformer_model, roformer_output_format, 
                 roformer_segment_size, roformer_override_segment_size, roformer_overlap, 
                 roformer_batch_size, roformer_normalization_threshold, roformer_amplification_threshold, 
                 roformer_single_stem], 
                [roformer_stem1, roformer_stem2]
            )

        with gr.TabItem("MDX23C"):
            with gr.Row():
                mdx23c_model = gr.Dropdown(
                    label="Select the model",
                    choices=mdx23c_models,
                    value=initial_settings.get("MDX23C", {}).get("model", None),
                    interactive=True
                )
                mdx23c_output_format = gr.Dropdown(
                    label="Select the output format",
                    choices=output_format,
                    value=initial_settings.get("MDX23C", {}).get("output_format", None),
                    interactive=True
                )
            
            with gr.Row():
                mdx23c_audio = gr.Audio(
                    label="Input audio",
                    type="filepath",
                    interactive=True
                )
            
            with gr.Row():
                mdx23c_button = gr.Button("Separate!", variant="primary")
            
            with gr.Row():
                mdx23c_stem1 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    label="Stem 1",
                    type="filepath"
                )
                mdx23c_stem2 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    label="Stem 2",
                    type="filepath"
                )
            
            with gr.Accordion("Separation by link", open=False):
                with gr.Row():
                    mdx23c_link = gr.Textbox(
                        label="Link",
                        placeholder="Paste the link here",
                        interactive=True
                    )
                with gr.Row():
                    gr.Markdown("You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)")
                with gr.Row():
                    mdx23c_download_button = gr.Button("Download!", variant="primary")

            mdx23c_download_button.click(download_audio, [mdx23c_link], [mdx23c_audio])

            with gr.Accordion("Batch separation", open=False):
                with gr.Row():
                    mdx23c_input_path = gr.Textbox(
                        label="Input path",
                        placeholder="Place the input path here",
                        interactive=True
                    )
                    mdx23c_output_path = gr.Textbox(
                        label="Output path",
                        placeholder="Place the output path here",
                        interactive=True
                    )
                with gr.Row():
                    mdx23c_bath_button = gr.Button("Separate!", variant="primary")
                with gr.Row():
                    mdx23c_info = gr.Textbox(
                        label="Output information",
                        interactive=False
                    )

        with gr.TabItem("MDX23C Processing"):
            with gr.Row():
                mdx23c_segment_size = gr.Slider(
                    minimum=32,
                    maximum=4000,
                    step=32,
                    label="Segment size",
                    info="Larger consumes more resources, but may give better results",
                    value=initial_settings.get("MDX23C", {}).get("segment_size", 256),
                    interactive=True
                )
                mdx23c_override_segment_size = gr.Checkbox(
                    label="Override segment size",
                    info="Override model default segment size instead of using the model default value",
                    value=initial_settings.get("MDX23C", {}).get("override_segment_size", False),
                    interactive=True
                )
            with gr.Row():
                mdx23c_overlap = gr.Slider(
                    minimum=2,
                    maximum=50,
                    step=1,
                    label="Overlap",
                    info="Amount of overlap between prediction windows",
                    value=initial_settings.get("MDX23C", {}).get("overlap", 8),
                    interactive=True
                )
                mdx23c_batch_size = gr.Slider(
                    label="Batch size",
                    info="Larger consumes more RAM but may process slightly faster",
                    minimum=1,
                    maximum=16,
                    step=1,
                    value=initial_settings.get("MDX23C", {}).get("batch_size", 1),
                    interactive=True
                )
            with gr.Row():
                mdx23c_normalization_threshold = gr.Slider(
                    label="Normalization threshold",
                    info="The threshold for audio normalization",
                    minimum=0.1,
                    maximum=1,
                    step=0.1,
                    value=initial_settings.get("MDX23C", {}).get("normalization_threshold", 0.9),
                    interactive=True
                )
                mdx23c_amplification_threshold = gr.Slider(
                    label="Amplification threshold",
                    info="The threshold for audio amplification",
                    minimum=0.1,
                    maximum=1,
                    step=0.1,
                    value=initial_settings.get("MDX23C", {}).get("amplification_threshold", 0.7),
                    interactive=True
                )
            with gr.Row():
                mdx23c_single_stem = gr.Textbox(
                    label="Output only single stem",
                    placeholder="Write the stem you want, check the stems of each model on Leaderboard. e.g. Instrumental",
                    value=initial_settings.get("MDX23C", {}).get("single_stem", ""),
                    interactive=True
                )

            mdx23c_bath_button.click(
                mdx23c_batch, 
                [mdx23c_input_path, mdx23c_output_path, mdx23c_model, mdx23c_output_format, 
                 mdx23c_segment_size, mdx23c_override_segment_size, mdx23c_overlap, 
                 mdx23c_batch_size, mdx23c_normalization_threshold, mdx23c_amplification_threshold, 
                 mdx23c_single_stem], 
                [mdx23c_info]
            )

            mdx23c_button.click(
                mdxc_separator, 
                [mdx23c_audio, mdx23c_model, mdx23c_output_format, 
                 mdx23c_segment_size, mdx23c_override_segment_size, mdx23c_overlap, 
                 mdx23c_batch_size, mdx23c_normalization_threshold, mdx23c_amplification_threshold, 
                 mdx23c_single_stem], 
                [mdx23c_stem1, mdx23c_stem2]
            )
                
        with gr.TabItem("MDX-NET"):
            with gr.Row():
                mdxnet_model = gr.Dropdown(
                    label="Select the model",
                    choices=mdxnet_models,
                    value=initial_settings.get("MDX-NET", {}).get("model", None),
                    interactive=True
                )
                mdxnet_output_format = gr.Dropdown(
                    label="Select the output format",
                    choices=output_format,
                    value=initial_settings.get("MDX-NET", {}).get("output_format", None),
                    interactive=True
                )
            
            with gr.Row():
                mdxnet_audio = gr.Audio(
                    label="Input audio",
                    type="filepath",
                    interactive=True
                )
            
            with gr.Row():
                mdxnet_button = gr.Button("Separate!", variant="primary")
            
            with gr.Row():
                mdxnet_stem1 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    label="Stem 1",
                    type="filepath"
                )
                mdxnet_stem2 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    label="Stem 2",
                    type="filepath"
                )
            
            with gr.Accordion("Separation by link", open=False):
                with gr.Row():
                    mdxnet_link = gr.Textbox(
                        label="Link",
                        placeholder="Paste the link here",
                        interactive=True
                    )
                with gr.Row():
                    gr.Markdown("You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)")
                with gr.Row():
                    mdxnet_download_button = gr.Button("Download!", variant="primary")
                
            mdxnet_download_button.click(download_audio, [mdxnet_link], [mdxnet_audio])

            with gr.Accordion("Batch separation", open=False):
                with gr.Row():
                    mdxnet_input_path = gr.Textbox(
                        label="Input path",
                        placeholder="Place the input path here",
                        interactive=True
                    )
                    mdxnet_output_path = gr.Textbox(
                        label="Output path",
                        placeholder="Place the output path here",
                        interactive=True
                    )
                with gr.Row():
                    mdxnet_bath_button = gr.Button("Separate!", variant="primary")
                with gr.Row():
                    mdxnet_info = gr.Textbox(
                        label="Output information",
                        interactive=False
                    )

        with gr.TabItem("MDX-NET Processing"):
            with gr.Row():
                mdxnet_hop_length = gr.Slider(
                    label="Hop length",
                    info="Usually called stride in neural networks; only change if you know what you're doing",
                    minimum=32,
                    maximum=2048,
                    step=32,
                    value=initial_settings.get("MDX-NET", {}).get("hop_length", 1024),
                    interactive=True
                )
                mdxnet_segment_size = gr.Slider(
                    minimum=32,
                    maximum=4000,
                    step=32,
                    label="Segment size",
                    info="Larger consumes more resources, but may give better results",
                    value=initial_settings.get("MDX-NET", {}).get("segment_size", 256),
                    interactive=True
                )
                mdxnet_denoise = gr.Checkbox(
                    label="Denoise",
                    info="Enable denoising during separation",
                    value=initial_settings.get("MDX-NET", {}).get("denoise", True),
                    interactive=True
                )
            with gr.Row():
                mdxnet_overlap = gr.Slider(
                    label="Overlap",
                    info="Amount of overlap between prediction windows",
                    minimum=0.001,
                    maximum=0.999,
                    step=0.001,
                    value=initial_settings.get("MDX-NET", {}).get("overlap", 0.25),
                    interactive=True
                )
                mdxnet_batch_size = gr.Slider(
                    label="Batch size",
                    info="Larger consumes more RAM but may process slightly faster",
                    minimum=1,
                    maximum=16,
                    step=1,
                    value=initial_settings.get("MDX-NET", {}).get("batch_size", 1),
                    interactive=True
                )
            with gr.Row():
                mdxnet_normalization_threshold = gr.Slider(
                    label="Normalization threshold",
                    info="The threshold for audio normalization",
                    minimum=0.1,
                    maximum=1,
                    step=0.1,
                    value=initial_settings.get("MDX-NET", {}).get("normalization_threshold", 0.9),
                    interactive=True
                )
                mdxnet_amplification_threshold = gr.Slider(
                    label="Amplification threshold",
                    info="The threshold for audio amplification",
                    minimum=0.1,
                    maximum=1,
                    step=0.1,
                    value=initial_settings.get("MDX-NET", {}).get("amplification_threshold", 0.7),
                    interactive=True
                )
            with gr.Row():
                mdxnet_single_stem = gr.Textbox(
                    label="Output only single stem",
                    placeholder="Write the stem you want, check the stems of each model on Leaderboard. e.g. Instrumental",
                    value=initial_settings.get("MDX-NET", {}).get("single_stem", ""),
                    interactive=True
                )

            mdxnet_bath_button.click(
                mdxnet_batch, 
                [mdxnet_input_path, mdxnet_output_path, mdxnet_model, mdxnet_output_format, 
                 mdxnet_hop_length, mdxnet_segment_size, mdxnet_denoise, mdxnet_overlap, 
                 mdxnet_batch_size, mdxnet_normalization_threshold, mdxnet_amplification_threshold, 
                 mdxnet_single_stem], 
                [mdxnet_info]
            )

            mdxnet_button.click(
                mdxnet_separator, 
                [mdxnet_audio, mdxnet_model, mdxnet_output_format, 
                 mdxnet_hop_length, mdxnet_segment_size, mdxnet_denoise, mdxnet_overlap, 
                 mdxnet_batch_size, mdxnet_normalization_threshold, mdxnet_amplification_threshold, 
                 mdxnet_single_stem], 
                [mdxnet_stem1, mdxnet_stem2]
            )

        with gr.TabItem("VR ARCH"):
            with gr.Row():
                vrarch_model = gr.Dropdown(
                    label="Select the model",
                    choices=vrarch_models,
                    value=initial_settings.get("VR Arch", {}).get("model", None),
                    interactive=True
                )
                vrarch_output_format = gr.Dropdown(
                    label="Select the output format",
                    choices=output_format,
                    value=initial_settings.get("VR Arch", {}).get("output_format", None),
                    interactive=True
                )
            
            with gr.Row():
                vrarch_audio = gr.Audio(
                    label="Input audio",
                    type="filepath",
                    interactive=True
                )
            
            with gr.Row():
                vrarch_button = gr.Button("Separate!", variant="primary")
            
            with gr.Row():
                vrarch_stem1 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    type="filepath",
                    label="Stem 1"
                )
                vrarch_stem2 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    type="filepath",
                    label="Stem 2"
                )
            
            with gr.Accordion("Separation by link", open=False):
                with gr.Row():
                    vrarch_link = gr.Textbox(
                        label="Link",
                        placeholder="Paste the link here",
                        interactive=True
                    )
                with gr.Row():
                    gr.Markdown("You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)")
                with gr.Row():
                    vrarch_download_button = gr.Button("Download!", variant="primary")

            vrarch_download_button.click(download_audio, [vrarch_link], [vrarch_audio])
            
            with gr.Accordion("Batch separation", open=False):
                with gr.Row():
                    vrarch_input_path = gr.Textbox(
                        label="Input path",
                        placeholder="Place the input path here",
                        interactive=True
                    )
                    vrarch_output_path = gr.Textbox(
                        label="Output path",
                        placeholder="Place the output path here",
                        interactive=True
                    )
                with gr.Row():
                    vrarch_bath_button = gr.Button("Separate!", variant="primary")
                with gr.Row():
                    vrarch_info = gr.Textbox(
                        label="Output information",
                        interactive=False
                    )

        with gr.TabItem("VR ARCH Processing"):
            with gr.Row():
                vrarch_window_size = gr.Slider(
                    label="Window size",
                    info="Balance quality and speed. 1024 = fast but lower, 320 = slower but better quality",
                    minimum=320,
                    maximum=1024,
                    step=32,
                    value=initial_settings.get("VR Arch", {}).get("window_size", 512),
                    interactive=True
                )
                vrarch_agression = gr.Slider(
                    minimum=1,
                    maximum=50,
                    step=1,
                    label="Agression",
                    info="Intensity of primary stem extraction",
                    value=initial_settings.get("VR Arch", {}).get("aggression", 5),
                    interactive=True
                )
                vrarch_tta = gr.Checkbox(
                    label="TTA",
                    info="Enable Test-Time-Augmentation; slow but improves quality",
                    value=initial_settings.get("VR Arch", {}).get("tta", True),
                    interactive=True
                )
            with gr.Row():
                vrarch_post_process = gr.Checkbox(
                    label="Post process",
                    info="Identify leftover artifacts within vocal output; may improve separation for some songs",
                    value=initial_settings.get("VR Arch", {}).get("post_process", False),
                    interactive=True
                )
                vrarch_post_process_threshold = gr.Slider(
                    label="Post process threshold",
                    info="Threshold for post-processing",
                    minimum=0.1,
                    maximum=0.3,
                    step=0.1,
                    value=initial_settings.get("VR Arch", {}).get("post_process_threshold", 0.2),
                    interactive=True
                )
            with gr.Row():
                vrarch_high_end_process = gr.Checkbox(
                    label="High end process",
                    info="Mirror the missing frequency range of the output",
                    value=initial_settings.get("VR Arch", {}).get("high_end_process", False),
                    interactive=True,
                )
                vrarch_batch_size = gr.Slider(
                    label="Batch size",
                    info="Larger consumes more RAM but may process slightly faster",
                    minimum=1,
                    maximum=16,
                    step=1,
                    value=initial_settings.get("VR Arch", {}).get("batch_size", 1),
                    interactive=True
                )
            with gr.Row():
                vrarch_normalization_threshold = gr.Slider(
                    label="Normalization threshold",
                    info="The threshold for audio normalization",
                    minimum=0.1,
                    maximum=1,
                    step=0.1,
                    value=initial_settings.get("VR Arch", {}).get("normalization_threshold", 0.9),
                    interactive=True
                )
                vrarch_amplification_threshold = gr.Slider(
                    label="Amplification threshold",
                    info="The threshold for audio amplification",
                    minimum=0.1,
                    maximum=1,
                    step=0.1,
                    value=initial_settings.get("VR Arch", {}).get("amplification_threshold", 0.7),
                    interactive=True
                )
            with gr.Row():
                vrarch_single_stem = gr.Textbox(
                    label="Output only single stem",
                    placeholder="Write the stem you want, check the stems of each model on Leaderboard. e.g. Instrumental",
                    value=initial_settings.get("VR Arch", {}).get("single_stem", ""),
                    interactive=True
                )

            vrarch_bath_button.click(
                vrarch_batch, 
                [vrarch_input_path, vrarch_output_path, vrarch_model, vrarch_output_format, 
                 vrarch_window_size, vrarch_agression, vrarch_tta, vrarch_post_process, 
                 vrarch_post_process_threshold, vrarch_high_end_process, vrarch_batch_size, 
                 vrarch_normalization_threshold, vrarch_amplification_threshold, vrarch_single_stem], 
                [vrarch_info]
            )

            vrarch_button.click(
                vrarch_separator, 
                [vrarch_audio, vrarch_model, vrarch_output_format, 
                 vrarch_window_size, vrarch_agression, vrarch_tta, vrarch_post_process, 
                 vrarch_post_process_threshold, vrarch_high_end_process, vrarch_batch_size, 
                 vrarch_normalization_threshold, vrarch_amplification_threshold, vrarch_single_stem], 
                [vrarch_stem1, vrarch_stem2]
            )

        with gr.TabItem("Demucs"):
            with gr.Row():
                demucs_model = gr.Dropdown(
                    label="Select the model",
                    choices=demucs_models,
                    value=initial_settings.get("Demucs", {}).get("model", None),
                    interactive=True
                )
                demucs_output_format = gr.Dropdown(
                    label="Select the output format",
                    choices=output_format,
                    value=initial_settings.get("Demucs", {}).get("output_format", None),
                    interactive=True
                )
            
            with gr.Row():
                demucs_audio = gr.Audio(
                    label="Input audio",
                    type="filepath",
                    interactive=True
                )
            
            with gr.Row():
                demucs_button = gr.Button("Separate!", variant="primary")
            
            with gr.Row():
                demucs_stem1 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    type="filepath",
                    label="Stem 1"
                )
                demucs_stem2 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    type="filepath",
                    label="Stem 2"
                )
            with gr.Row():
                demucs_stem3 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    type="filepath",
                    label="Stem 3"
                )
                demucs_stem4 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    type="filepath",
                    label="Stem 4"
                )
            with gr.Row(visible=False) as stem6:
                demucs_stem5 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    type="filepath",
                    label="Stem 5"
                )
                demucs_stem6 = gr.Audio(
                    show_download_button=True,
                    interactive=False,
                    type="filepath",
                    label="Stem 6"
                )

            demucs_model.change(update_stems, inputs=[demucs_model], outputs=stem6)
            
            with gr.Accordion("Separation by link", open=False):
                with gr.Row():
                    demucs_link = gr.Textbox(
                        label="Link",
                        placeholder="Paste the link here",
                        interactive=True
                    )
                with gr.Row():
                    gr.Markdown("You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)")
                with gr.Row():
                    demucs_download_button = gr.Button("Download!", variant="primary")

            demucs_download_button.click(download_audio, [demucs_link], [demucs_audio])

            with gr.Accordion("Batch separation", open=False):
                with gr.Row():
                    demucs_input_path = gr.Textbox(
                        label="Input path",
                        placeholder="Place the input path here",
                        interactive=True
                    )
                    demucs_output_path = gr.Textbox(
                        label="Output path",
                        placeholder="Place the output path here",
                        interactive=True
                    )
                with gr.Row():
                    demucs_bath_button = gr.Button("Separate!", variant="primary")
                with gr.Row():
                    demucs_info = gr.Textbox(
                        label="Output information",
                        interactive=False
                    )

        with gr.TabItem("Demucs Processing"):
            with gr.Row():
                demucs_shifts = gr.Slider(
                    label="Shifts",
                    info="Number of predictions with random shifts, higher = slower but better quality",
                    minimum=1,
                    maximum=20,
                    step=1,
                    value=initial_settings.get("Demucs", {}).get("shifts", 2),
                    interactive=True
                )
                demucs_segment_size = gr.Slider(
                    label="Segment size",
                    info="Size of segments into which the audio is split. Higher = slower but better quality",
                    minimum=1,
                    maximum=100,
                    step=1,
                    value=initial_settings.get("Demucs", {}).get("segment_size", 40),
                    interactive=True
                )
                demucs_segments_enabled = gr.Checkbox(
                    label="Segment-wise processing",
                    info="Enable segment-wise processing",
                    value=initial_settings.get("Demucs", {}).get("segments_enabled", True),
                    interactive=True
                )
            with gr.Row():
                demucs_overlap = gr.Slider(
                    label="Overlap",
                    info="Overlap between prediction windows. Higher = slower but better quality",
                    minimum=0.001,
                    maximum=0.999,
                    step=0.001,
                    value=initial_settings.get("Demucs", {}).get("overlap", 0.25),
                    interactive=True
                )
                demucs_batch_size = gr.Slider(
                    label="Batch size",
                    info="Larger consumes more RAM but may process slightly faster",
                    minimum=1,
                    maximum=16,
                    step=1,
                    value=initial_settings.get("Demucs", {}).get("batch_size", 1),
                    interactive=True
                )
            with gr.Row():
                demucs_normalization_threshold = gr.Slider(
                    label="Normalization threshold",
                    info="The threshold for audio normalization",
                    minimum=0.1,
                    maximum=1,
                    step=0.1,
                    value=initial_settings.get("Demucs", {}).get("normalization_threshold", 0.9),
                    interactive=True
                )
                demucs_amplification_threshold = gr.Slider(
                    label="Amplification threshold",
                    info="The threshold for audio amplification",
                    minimum=0.1,
                    maximum=1,
                    step=0.1,
                    value=initial_settings.get("Demucs", {}).get("amplification_threshold", 0.7),
                    interactive=True
                )

            demucs_bath_button.click(
                demucs_batch, 
                [demucs_input_path, demucs_output_path, demucs_model, demucs_output_format, 
                 demucs_shifts, demucs_segment_size, demucs_segments_enabled, demucs_overlap, 
                 demucs_batch_size, demucs_normalization_threshold, demucs_amplification_threshold], 
                [demucs_info]
            )
                
            demucs_button.click(
                demucs_separator, 
                [demucs_audio, demucs_model, demucs_output_format, 
                 demucs_shifts, demucs_segment_size, demucs_segments_enabled, demucs_overlap, 
                 demucs_batch_size, demucs_normalization_threshold, demucs_amplification_threshold], 
                [demucs_stem1, demucs_stem2, demucs_stem3, demucs_stem4, demucs_stem5, demucs_stem6]
            )

        with gr.TabItem("Leaderboard"):
            with gr.Group():
                with gr.Row(equal_height=True):
                    list_filter = gr.Dropdown(
                        label="List filter",
                        info="Filter and sort the model list by stem",
                        choices=["vocals", "instrumental", "reverb", "echo", "noise", "crowd", "dry", "aspiration", "male", "woodwinds", "kick", "drums", "bass", "guitar", "piano", "other"],
                        value=lambda: None
                    )
                    list_button = gr.Button("Show list!", variant="primary")
            output_list = gr.HTML(label="Leaderboard")

            list_button.click(leaderboard, inputs=list_filter, outputs=output_list)

        with gr.TabItem("Credits"):
            gr.Markdown(
                """
                UVR5 UI created by **[Eddycrack 864](https://github.com/Eddycrack864). Improved for HF only by [BF667](https://github.com/BF667)
                * python-audio-separator by [beveradb](https://github.com/beveradb).
                * Thanks to [Mikus](https://github.com/cappuch) for the help with the code.
                * Thanks to [Nick088](https://huggingface.co/Nick088) for the help to fix roformers.
                * Thanks to [yt_dlp](https://github.com/yt-dlp/yt-dlp) devs.
                * Thanks to [Bebra777228](https://github.com/Bebra777228)'s code for guiding me to improve my code.
                * Thanks to Nick088, MrM0dZ, BF667, lucinamari, perariroswe, Enes, Léo and the_undead0 for helping translate UVR5 UI.
                * Thanks to vadigr123 for creating the images for the Discord Rich Presence.
                
                You can donate to the original UVR5 project here:
                [!["Buy Me A Coffee"](https://www.buymeacoffee.com/assets/img/custom_images/orange_img.png)](https://www.buymeacoffee.com/uvr5)
                """
            )

app.queue()
app.launch()