{
  "cells": [
      {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/w-okada/voice-changer/blob/master/Realtime_Voice_Changer_on_Colab.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Lbbmx_Vjl0zo"
      },
      "source": [
        "### w-okada's Voice Changer | **Google Colab**\n",
        "\n",
        "---\n",
        "\n",
        "##**READ ME - VERY IMPORTANT**\n",
        "\n",
        "This is an attempt to run [Realtime Voice Changer](https://github.com/w-okada/voice-changer) on Google Colab, still not perfect but is totally usable, you can use the following settings for better results:\n",
        "\n",
        "If you're using a index: `f0: RMVPE_ONNX | Chunk: 112 or higher | Extra: 8192`\\\n",
        "If you're not using a index: `f0: RMVPE_ONNX | Chunk: 96 or higher | Extra: 16384`\\\n",
        "**Don't forget to select your Colab GPU in the GPU field (<b>Tesla T4</b>, for free users)*\n",
        "> Seems that PTH models performance better than ONNX for now, you can still try ONNX models and see if it satisfies you\n",
        "\n",
        "\n",
        "*You can always [click here](https://github.com/YunaOneeChan/Voice-Changer-Settings) to check if these settings are up-to-date*\n",
        "<br><br>\n",
        "\n",
        "---\n",
        "\n",
        "###Always use Colab GPU (**VERY VERY VERY IMPORTANT!**)\n",
        "You need to use a Colab GPU so the Voice Changer can work faster and better\\\n",
        "Use the menu above and click on **Runtime** » **Change runtime** » **Hardware acceleration** to select a GPU (**T4 is the free one**)\n",
        "\n",
        "---\n",
        "\n",
        "<br>\n",
        "\n",
        "# **Credits and Support**\n",
        "Realtime Voice Changer by [w-okada](https://github.com/w-okada)\\\n",
        "Colab files updated by [rafacasari](https://github.com/Rafacasari)\\\n",
        "Recommended settings by [YunaOneeChan](https://github.com/YunaOneeChan)\n",
        "\n",
        "Need help? [AI Hub Discord](https://discord.gg/aihub) » ***#help-realtime-vc***\n",
        "\n",
        "---"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "cellView": "form",
        "id": "RhdqDSt-LfGr"
      },
      "outputs": [],
      "source": [
        "# @title **[Optional]** Connect to Google Drive\n",
        "# @markdown Using Google Drive can improve load times a bit and your models will be stored, so you don't need to re-upload every time that you use.\n",
        "import os\n",
        "from google.colab import drive\n",
        "\n",
        "if not os.path.exists('/content/drive'):\n",
        "  drive.mount('/content/drive')\n",
        "\n",
        "%cd /content/drive/MyDrive"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "86wTFmqsNMnD",
        "cellView": "form"
      },
      "outputs": [],
      "source": [
        "# @title **[1]** Clone repository and install dependencies\n",
        "# @markdown This first step will download the latest version of Voice Changer and install the dependencies. **It will take around 2 minutes to complete.**\n",
        "\n",
        "!git clone --depth 1 https://github.com/w-okada/voice-changer.git &> /dev/null\n",
        "\n",
        "%cd voice-changer/server/\n",
        "print(\"\\033[92mSuccessfully cloned the repository\")\n",
        "\n",
        "!apt-get install libportaudio2 &> /dev/null\n",
	"!pip install pyworld"
        "!pip install onnxruntime-gpu uvicorn faiss-gpu fairseq jedi google-colab moviepy decorator==4.4.2 sounddevice numpy==1.23.5 pyngrok --quiet\n",
        "!pip install -r requirements.txt --no-build-isolation --quiet\n",
        "# Maybe install Tensor packages?\n",
        "#!pip install torch-tensorrt\n",
        "#!pip install TensorRT\n",
        "print(\"\\033[92mSuccessfully installed all packages!\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "lLWQuUd7WW9U",
        "cellView": "form"
      },
      "outputs": [],
      "source": [
        "# @title **[2]** Start Server **using ngrok** (Recommended | **need a ngrok account**)\n",
        "# @markdown This cell will start the server, the first time that you run it will download the models, so it can take a while (~1-2 minutes)\n",
        "\n",
        "# @markdown ---\n",
        "# @markdown You'll need a ngrok account, but **it's free**!\n",
        "# @markdown ---\n",
        "# @markdown **1** - Create a **free** account at [ngrok](https://dashboard.ngrok.com/signup)\\\n",
        "# @markdown **2** - If you didn't logged in with Google or Github, you will need to **verify your e-mail**!\\\n",
        "# @markdown **3** - Click [this link](https://dashboard.ngrok.com/get-started/your-authtoken) to get your auth token, copy it and place it here:\n",
        "from pyngrok import conf, ngrok\n",
        "\n",
        "Token = '' # @param {type:\"string\"}\n",
        "# @markdown **4** - Still need further tests, but maybe region can help a bit on latency?\\\n",
        "# @markdown `Default Region: us - United States (Ohio)`\n",
        "Region = \"us - United States (Ohio)\" # @param [\"ap - Asia/Pacific (Singapore)\", \"au - Australia (Sydney)\",\"eu - Europe (Frankfurt)\", \"in - India (Mumbai)\",\"jp - Japan (Tokyo)\",\"sa - South America (Sao Paulo)\", \"us - United States (Ohio)\"]\n",
        "\n",
        "MyConfig = conf.PyngrokConfig()\n",
        "\n",
        "MyConfig.auth_token = Token\n",
        "MyConfig.region = Region[0:2]\n",
        "\n",
        "conf.get_default().authtoken = Token\n",
        "conf.get_default().region = Region[0:2]\n",
        "\n",
        "conf.set_default(MyConfig);\n",
        "\n",
        "# @markdown ---\n",
        "# @markdown If you want to automatically clear the output when the server loads, check this option.\n",
        "Clear_Output = True # @param {type:\"boolean\"}\n",
        "\n",
        "import portpicker, subprocess, threading, time, socket, urllib.request\n",
        "PORT = portpicker.pick_unused_port()\n",
        "\n",
        "from IPython.display import clear_output, Javascript\n",
        "\n",
        "from pyngrok import ngrok\n",
        "ngrokConnection = ngrok.connect(PORT)\n",
        "public_url = ngrokConnection.public_url\n",
        "\n",
        "def iframe_thread(port):\n",
        "  while True:\n",
        "      time.sleep(0.5)\n",
        "      sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)\n",
        "      result = sock.connect_ex(('127.0.0.1', port))\n",
        "      if result == 0:\n",
        "        break\n",
        "      sock.close()\n",
        "  clear_output()\n",
        "  print(\"------- SERVER READY! -------\")\n",
        "  print(\"Your server is available at:\")\n",
        "  print(public_url)\n",
        "  print(\"-----------------------------\")\n",
        "  display(Javascript('window.open(\"{url}\", \\'_blank\\');'.format(url=public_url)))\n",
        "\n",
        "threading.Thread(target=iframe_thread, daemon=True, args=(PORT,)).start()\n",
        "\n",
        "!python3 MMVCServerSIO.py \\\n",
        "  -p {PORT} \\\n",
        "  --https False \\\n",
        "  --content_vec_500 pretrain/checkpoint_best_legacy_500.pt \\\n",
        "  --content_vec_500_onnx pretrain/content_vec_500.onnx \\\n",
        "  --content_vec_500_onnx_on true \\\n",
        "  --hubert_base pretrain/hubert_base.pt \\\n",
        "  --hubert_base_jp pretrain/rinna_hubert_base_jp.pt \\\n",
        "  --hubert_soft pretrain/hubert/hubert-soft-0d54a1f4.pt \\\n",
        "  --nsf_hifigan pretrain/nsf_hifigan/model \\\n",
        "  --crepe_onnx_full pretrain/crepe_onnx_full.onnx \\\n",
        "  --crepe_onnx_tiny pretrain/crepe_onnx_tiny.onnx \\\n",
        "  --rmvpe pretrain/rmvpe.pt \\\n",
        "  --model_dir model_dir \\\n",
        "  --samples samples.json"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# @title **[Optional]** Start Server **using localtunnel** (ngrok alternative | no account needed)\n",
        "# @markdown This cell will start the server, the first time that you run it will download the models, so it can take a while (~1-2 minutes)\n",
        "\n",
        "# @markdown ---\n",
        "!npm config set update-notifier false\n",
        "!npm install -g localtunnel\n",
        "print(\"\\033[92mLocalTunnel installed!\")\n",
        "# @markdown If you want to automatically clear the output when the server loads, check this option.\n",
        "Clear_Output = True # @param {type:\"boolean\"}\n",
        "\n",
        "import portpicker, subprocess, threading, time, socket, urllib.request\n",
        "PORT = portpicker.pick_unused_port()\n",
        "\n",
        "from IPython.display import clear_output, Javascript\n",
        "\n",
        "def iframe_thread(port):\n",
        "  while True:\n",
        "      time.sleep(0.5)\n",
        "      sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)\n",
        "      result = sock.connect_ex(('127.0.0.1', port))\n",
        "      if result == 0:\n",
        "        break\n",
        "      sock.close()\n",
        "  clear_output()\n",
        "  print(\"Use the following endpoint to connect to localtunnel:\", urllib.request.urlopen('https://ipv4.icanhazip.com').read().decode('utf8').strip(\"\\n\"))\n",
        "  p = subprocess.Popen([\"lt\", \"--port\", \"{}\".format(port)], stdout=subprocess.PIPE)\n",
        "  for line in p.stdout:\n",
        "    print(line.decode(), end='')\n",
        "\n",
        "threading.Thread(target=iframe_thread, daemon=True, args=(PORT,)).start()\n",
        "\n",
        "\n",
        "!python3 MMVCServerSIO.py \\\n",
        "  -p {PORT} \\\n",
        "  --https False \\\n",
        "  --content_vec_500 pretrain/checkpoint_best_legacy_500.pt \\\n",
        "  --content_vec_500_onnx pretrain/content_vec_500.onnx \\\n",
        "  --content_vec_500_onnx_on true \\\n",
        "  --hubert_base pretrain/hubert_base.pt \\\n",
        "  --hubert_base_jp pretrain/rinna_hubert_base_jp.pt \\\n",
        "  --hubert_soft pretrain/hubert/hubert-soft-0d54a1f4.pt \\\n",
        "  --nsf_hifigan pretrain/nsf_hifigan/model \\\n",
        "  --crepe_onnx_full pretrain/crepe_onnx_full.onnx \\\n",
        "  --crepe_onnx_tiny pretrain/crepe_onnx_tiny.onnx \\\n",
        "  --rmvpe pretrain/rmvpe.pt \\\n",
        "  --model_dir model_dir \\\n",
        "  --samples samples.json \\\n",
        "  --colab True"
      ],
      "metadata": {
        "cellView": "form",
        "id": "ZwZaCf4BeZi2"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# In Development | **Need contributors**"
      ],
      "metadata": {
        "id": "iuf9pBHYpTn-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# @title **[BROKEN]** Start Server using Colab Tunnels (trying to fix this TwT)\n",
        "# @markdown **Issue:** Everything starts correctly, but when you try to use the client, you'll see in your browser console a bunch of errors **(Error 500 - Not Allowed.)**\n",
        "\n",
        "import portpicker, subprocess, threading, time, socket, urllib.request\n",
        "PORT = portpicker.pick_unused_port()\n",
        "\n",
        "def iframe_thread(port):\n",
        "  while True:\n",
        "      time.sleep(0.5)\n",
        "      sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)\n",
        "      result = sock.connect_ex(('127.0.0.1', port))\n",
        "      if result == 0:\n",
        "        break\n",
        "      sock.close()\n",
        "  from google.colab.output import serve_kernel_port_as_window\n",
        "  serve_kernel_port_as_window(PORT)\n",
        "\n",
        "threading.Thread(target=iframe_thread, daemon=True, args=(PORT,)).start()\n",
        "\n",
        "!python3 MMVCServerSIO.py \\\n",
        "  -p {PORT} \\\n",
        "  --https False \\\n",
        "  --content_vec_500 pretrain/checkpoint_best_legacy_500.pt \\\n",
        "  --content_vec_500_onnx pretrain/content_vec_500.onnx \\\n",
        "  --content_vec_500_onnx_on true \\\n",
        "  --hubert_base pretrain/hubert_base.pt \\\n",
        "  --hubert_base_jp pretrain/rinna_hubert_base_jp.pt \\\n",
        "  --hubert_soft pretrain/hubert/hubert-soft-0d54a1f4.pt \\\n",
        "  --nsf_hifigan pretrain/nsf_hifigan/model \\\n",
        "  --crepe_onnx_full pretrain/crepe_onnx_full.onnx \\\n",
        "  --crepe_onnx_tiny pretrain/crepe_onnx_tiny.onnx \\\n",
        "  --rmvpe pretrain/rmvpe.pt \\\n",
        "  --model_dir model_dir \\\n",
        "  --samples samples.json"
      ],
      "metadata": {
        "id": "P2BN-iWvDrMM",
        "cellView": "form"
      },
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "colab": {
      "provenance": [],
      "private_outputs": true,
	  "include_colab_link": true,
      "gpuType": "T4",
      "collapsed_sections": [
        "iuf9pBHYpTn-"
      ]
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "nbformat": 4,
  "nbformat_minor": 0
}