{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "from pathlib import Path\n",
        "\n",
        "repo_path = Path(\"/content/AudioSep\")\n",
        "if not repo_path.exists():\n",
        "    !git clone https://github.com/Audio-AGI/AudioSep.git\n",
        "\n",
        "%cd /content/AudioSep"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "pjIhw5ECS_3_"
      },
      "outputs": [],
      "source": [
        "!pip install torchlibrosa==0.1.0 gradio==3.47.1 gdown lightning transformers==4.28.1 ftfy braceexpand webdataset soundfile wget h5py"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "t6h9KB3CcjBd"
      },
      "outputs": [],
      "source": [
        "checkpoints_dir = Path(\"checkpoint\")\n",
        "checkpoints_dir.mkdir(exist_ok=True)\n",
        "\n",
        "models = (\n",
        "    (\n",
        "        \"https://huggingface.co/spaces/badayvedat/AudioSep/resolve/main/checkpoint/audiosep_base_4M_steps.ckpt\",\n",
        "        checkpoints_dir / \"audiosep_base_4M_steps.ckpt\"\n",
        "    ),\n",
        "    (\n",
        "        \"https://huggingface.co/spaces/badayvedat/AudioSep/resolve/main/checkpoint/music_speech_audioset_epoch_15_esc_89.98.pt\",\n",
        "        checkpoints_dir / \"music_speech_audioset_epoch_15_esc_89.98.pt\"\n",
        "    )\n",
        ")\n",
        "\n",
        "for model_url, model_path in models:\n",
        "    if not model_path.exists():\n",
        "        !wget {model_url} -O {model_path}"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "3uDrzCQyY58h"
      },
      "outputs": [],
      "source": [
        "!wget \"https://audio-agi.github.io/Separate-Anything-You-Describe/demos/exp31_water drops_mixture.wav\""
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "0nr77CGXTwO1"
      },
      "outputs": [],
      "source": [
        "import torch\n",
        "from pipeline import build_audiosep, inference\n",
        "\n",
        "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
        "\n",
        "model = build_audiosep(\n",
        "      config_yaml='config/audiosep_base.yaml',\n",
        "      checkpoint_path=str(models[0][1]),\n",
        "      device=device)\n",
        "\n",
        "audio_file = 'exp31_water drops_mixture.wav'\n",
        "text = 'water drops'\n",
        "output_file='separated_audio.wav'\n",
        "\n",
        "# AudioSep processes the audio at 32 kHz sampling rate\n",
        "inference(model, audio_file, text, output_file, device)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "kssOe0pbPSWp"
      },
      "outputs": [],
      "source": [
        "print(f\"The separated audio is saved to: '{output_file}' file.\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "sl35U3dAR6KN"
      },
      "outputs": [],
      "source": []
    }
  ],
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}