{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import torch\n",
    "import sovits.sovits_inferencer\n",
    "from pydub import AudioSegment\n",
    "\n",
    "import IPython.display as ipd\n",
    "\n",
    "sovits_inferencer = sovits.sovits_inferencer.SovitsInferencer(\"sovits/configs/hoshimi_base.json\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "input_file = \"干花.mp3\"\n",
    "trans = 0\n",
    "slice_db = -40\n",
    "\n",
    "\n",
    "input_dir_path = \"inputs\"\n",
    "output_dir_path = \"outputs\"\n",
    "ext = input_file.split(\".\")[-1]\n",
    "input_path = os.path.join(input_dir_path, input_file)\n",
    "input_name = os.path.basename(input_path).replace(f\".{ext}\", \"\")\n",
    "\n",
    "vocals_path = os.path.join(output_dir_path, \"htdemucs\", input_name, \"vocals.wav\")\n",
    "no_vocals_path = os.path.join(output_dir_path, \"htdemucs\", input_name, \"no_vocals.wav\")\n",
    "dst_vocals_path = os.path.join(output_dir_path, f\"{input_name}_vocals_{trans}.wav\")\n",
    "output_path = os.path.join(output_dir_path, f\"{input_name}_{trans}.wav\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.makedirs(output_dir_path, exist_ok=True)\n",
    "os.system(f'python -m demucs.separate -d {\"cuda\" if torch.cuda.is_available() else \"cpu\"} --two-stems=vocals {input_path} -o {output_dir_path}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ipd.display(ipd.Audio(vocals_path))\n",
    "ipd.display(ipd.Audio(no_vocals_path))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from gradio.processing_utils import audio_to_file\n",
    "\n",
    "msg, (sr, audio) = sovits_inferencer.infer(vocals_path, None, trans, slice_db, 1000)\n",
    "print(msg)\n",
    "audio_to_file(sr, audio, dst_vocals_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ipd.display(ipd.Audio(dst_vocals_path))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def combine(vocals_path, no_vocals_path, output_path):\n",
    "    sound1 = AudioSegment.from_wav(vocals_path)\n",
    "    sound2 = AudioSegment.from_wav(no_vocals_path)\n",
    "\n",
    "    combined = sound1.overlay(sound2).set_frame_rate(sound1.frame_rate)\n",
    "    combined.export(output_path, format=\"wav\")\n",
    "    return combined\n",
    "\n",
    "combine(dst_vocals_path, no_vocals_path, output_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ipd.display(ipd.Audio(output_path))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.13"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "3d236089fc20612273a7c4c44c2149c93909997a9c6abd716b8f4ac89b8b9538"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}