| import gradio as gr
|
| import os, sys, subprocess
|
| import tempfile
|
| from scipy import signal
|
| import numpy as np
|
| from datetime import datetime
|
|
|
| if not __package__:
|
| from audio import Audio
|
| else:
|
| from ..audio import Audio
|
|
|
|
|
| class Plugin(Audio):
|
| def __init__(self):
|
| super().__init__()
|
| self.name = "Вычитание фантомного центра"
|
| self.requirements = []
|
| self.install_requirements(self.requirements)
|
| self.w_types = [
|
| "boxcar",
|
| "triang",
|
| "blackman",
|
| "hamming",
|
| "hann",
|
| "bartlett",
|
| "flattop",
|
| "parzen",
|
| "bohman",
|
| "blackmanharris",
|
| "nuttall",
|
| "barthann",
|
| "cosine",
|
| "exponential",
|
| "tukey",
|
| "taylor",
|
| "lanczos",
|
| ]
|
|
|
| def install_requirements(self, requirements: list):
|
| if requirements:
|
| cmd = [os.sys.executable, "-m", "pip", "install"]
|
| for pkg in requirements:
|
| cmd.append(pkg)
|
| result = subprocess.run(cmd, text=True, capture_output=True)
|
|
|
| def remove_center(
|
| self,
|
| input_file,
|
| output_format="flac",
|
| out_center="center.flac",
|
| out_stereo_base="stereo_base.flac",
|
| rdf=0.99999,
|
| window_size=4096,
|
| overlap=2,
|
| window_type="blackman",
|
| stereo_mode="stereo",
|
| ):
|
| output_file = out_stereo_base
|
| output_center_file = out_center
|
| data, samplerate, _ = self.read(i=input_file, mono=False, sr=None)
|
|
|
| if data.ndim != 2 or data.shape[0] != 2:
|
| raise ValueError("Требуется стереофайл (2 канала)")
|
|
|
| left = data[0, :]
|
| right = data[1, :]
|
| mono = left * 0.5 + right * 0.5
|
|
|
| nperseg = window_size
|
| noverlap = nperseg // overlap
|
|
|
| f, t, Z_left = signal.stft(
|
| left, fs=samplerate, nperseg=nperseg, noverlap=noverlap, window=window_type
|
| )
|
| f, t, Z_right = signal.stft(
|
| right, fs=samplerate, nperseg=nperseg, noverlap=noverlap, window=window_type
|
| )
|
| f, t, Z_mono = signal.stft(
|
| mono, fs=samplerate, nperseg=nperseg, noverlap=noverlap, window=window_type
|
| )
|
| if stereo_mode == "mono":
|
| Z_common_left = np.minimum(np.abs(Z_left), np.abs(Z_right)) * np.exp(
|
| 1j * np.angle(Z_mono)
|
| )
|
| Z_common_right = np.minimum(np.abs(Z_left), np.abs(Z_right)) * np.exp(
|
| 1j * np.angle(Z_mono)
|
| )
|
| else:
|
| Z_common_left = np.minimum(np.abs(Z_left), np.abs(Z_right)) * np.exp(
|
| 1j * np.angle(Z_right)
|
| )
|
| Z_common_right = np.minimum(np.abs(Z_left), np.abs(Z_right)) * np.exp(
|
| 1j * np.angle(Z_left)
|
| )
|
|
|
| reduction_factor = rdf
|
|
|
| Z_new_left = Z_left - Z_common_left * reduction_factor
|
| Z_new_right = Z_right - Z_common_right * reduction_factor
|
|
|
| _, new_left = signal.istft(
|
| Z_new_left,
|
| fs=samplerate,
|
| nperseg=nperseg,
|
| noverlap=noverlap,
|
| window=window_type,
|
| )
|
| _, new_right = signal.istft(
|
| Z_new_right,
|
| fs=samplerate,
|
| nperseg=nperseg,
|
| noverlap=noverlap,
|
| window=window_type,
|
| )
|
|
|
| _, common_signal_left = signal.istft(
|
| Z_common_left,
|
| fs=samplerate,
|
| nperseg=nperseg,
|
| noverlap=noverlap,
|
| window=window_type,
|
| )
|
| _, common_signal_right = signal.istft(
|
| Z_common_right,
|
| fs=samplerate,
|
| nperseg=nperseg,
|
| noverlap=noverlap,
|
| window=window_type,
|
| )
|
|
|
| new_left = new_left[: len(left)]
|
| new_right = new_right[: len(right)]
|
| common_signal_left = common_signal_left[: len(left)]
|
| common_signal_right = common_signal_right[: len(right)]
|
|
|
| peak = np.max([np.abs(new_left).max(), np.abs(new_right).max()])
|
| if peak > 1.0:
|
| new_left = new_left / peak
|
| new_right = new_right / peak
|
|
|
| output_file = self.write(
|
| o=output_file,
|
| array=np.column_stack((new_left, new_right)),
|
| sr=samplerate,
|
| of=output_format,
|
| br="320k",
|
| )
|
|
|
| inverted_center_left = -common_signal_left
|
| inverted_center_right = -common_signal_right
|
|
|
| mixed_left = left + inverted_center_left
|
| mixed_right = right + inverted_center_right
|
|
|
| peak_mixed = np.max([np.abs(mixed_left).max(), np.abs(mixed_right).max()])
|
| if peak_mixed > 1.0:
|
| mixed_left = mixed_left / peak_mixed
|
| mixed_right = mixed_right / peak_mixed
|
|
|
| output_center_file = self.write(
|
| o=output_center_file,
|
| array=np.column_stack((common_signal_left, common_signal_right)),
|
| sr=samplerate,
|
| of=output_format,
|
| br="320k",
|
| )
|
|
|
| return (output_file, output_center_file)
|
|
|
| def UI(self):
|
| with gr.Row():
|
| rmv_center_ui_input_audio = gr.File(
|
| label="Входное аудио",
|
| interactive=True,
|
| type="filepath",
|
| file_count="single",
|
| file_types=[f".{of}" for of in self.input_formats],
|
| )
|
| with gr.Group():
|
| with gr.Row():
|
| rmv_center_ui_reduction_f = gr.Slider(
|
| 0.1,
|
| 10,
|
| value=1,
|
| step=0.1,
|
| label="Фактор подавления",
|
| interactive=True,
|
| visible=False,
|
| )
|
| rmv_center_ui_overlap = gr.Slider(
|
| 2,
|
| 30,
|
| value=2,
|
| step=1,
|
| label="Перекрытие",
|
| interactive=True,
|
| visible=True,
|
| )
|
| rmv_center_ui_window_size = gr.Number(
|
| label="Размер окна",
|
| interactive=True,
|
| visible=True,
|
| minimum=32,
|
| maximum=882000,
|
| precision=1,
|
| value=2048,
|
| )
|
| with gr.Row():
|
| rmv_center_ui_format = gr.Dropdown(
|
| self.output_formats,
|
| value=self.output_formats[0],
|
| filterable=False,
|
| label="Формат выходного файла",
|
| interactive=True,
|
| )
|
| rmv_center_ui_window_types = gr.Dropdown(
|
| self.w_types,
|
| value=self.w_types[4],
|
| filterable=False,
|
| label="Тип окна",
|
| interactive=True,
|
| )
|
|
|
| rmv_center_ui_mono_mode = gr.Dropdown(
|
| ["mono", "stereo"],
|
| value="mono",
|
| filterable=False,
|
| label="Стерео-режим",
|
| interactive=True,
|
| )
|
| rmv_center_ui_extract_btn = gr.Button("Разделить")
|
| with gr.Group():
|
| with gr.Column():
|
| with gr.Row():
|
| rmv_center_ui_mid = gr.Audio(
|
| type="filepath",
|
| interactive=False,
|
| label="Фантомный центр",
|
| visible=True,
|
| show_download_button=True,
|
| )
|
| rmv_center_ui_side = gr.Audio(
|
| type="filepath",
|
| interactive=False,
|
| label="Стерео-база",
|
| visible=True,
|
| show_download_button=True,
|
| )
|
|
|
| @rmv_center_ui_extract_btn.click(
|
| inputs=[
|
| rmv_center_ui_input_audio,
|
| rmv_center_ui_format,
|
| rmv_center_ui_reduction_f,
|
| rmv_center_ui_window_size,
|
| rmv_center_ui_overlap,
|
| rmv_center_ui_window_types,
|
| rmv_center_ui_mono_mode,
|
| ],
|
| outputs=[rmv_center_ui_side, rmv_center_ui_mid],
|
| )
|
| def wrap_remove_center(input_audio, output_format, rf, ws, ovlp, wt, mono_mode):
|
| if input_audio:
|
| temp_dir = tempfile.mkdtemp(prefix="remove_center_")
|
| basename = self.short(
|
| os.path.splitext(os.path.basename(input_audio))[0], length=80
|
| )
|
| side, mid = self.remove_center(
|
| input_file=input_audio,
|
| output_format=output_format,
|
| out_center=os.path.join(
|
| temp_dir, f"{basename}_center.{output_format}"
|
| ),
|
| out_stereo_base=os.path.join(
|
| temp_dir, f"{basename}_stereo_base.{output_format}"
|
| ),
|
| overlap=ovlp,
|
| rdf=rf,
|
| stereo_mode=mono_mode,
|
| window_size=ws,
|
| window_type=wt,
|
| )
|
| return side, mid
|
|
|
