FaceFusion_UI.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/GitCorion/Elcano/blob/main/FaceFusion_UI.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hgVreYca3LcQ"
      },
      "source": [
        "# **FaceFusion WEB UI**\n",
        "\n",
        "DeepFake AI Tool\n",
        "\n",
        "**WARNING: COULD RISK YOUR FREE TIER COLAB ACC**\n",
        "\n",
        "Credits:\n",
        "\n",
        "- [Nick088](https://linktr.ee/Nick088) (now taking support of the [FaceFusion Unofficial Online Ports](https://github.com/Nick088Official/FaceFusion-Online-Ports).)\n",
        "- [hina](https://linktr.ee/_hina__) (simplified the code that makes the public gradio link appear, putting share=True without also messing up anything if you cancel execution and re run it.)\n",
        "- [Original FaceFusion team](https://github.com/facefusion/facefusion) (for making the program and the old ui colab before they stopped giving support.)\n",
        "\n",
        "## BASIC GUIDE:\n",
        "1. Choose a runtime type with Runtime -> Change runtime type (CPU is slower, any GPUs like the T4 free daily GPU are faster), Run the Install FaceFusion cell and wait until it finishes.\n",
        "2. (OPTIONAL) Run the Mount Google Drive cell, by default it will make your Google Drive FF_Colab/Outputs and use it as the output path.\n",
        "3. **(ONLY FOR WHO MOUNTED DRIVE)** Now you got 2 ways to upload the inputs:\n",
        "\n",
        "**For Google Chrome Users:**\n",
        "1. Run the Mount Google Drive Cell\n",
        "\n",
        "2. Run the 'Upload Target Vid/Pic & Source Pic' cell and upload both of them (either do ctrl and select both of them or just select one, then run it again, and select the other)\n",
        "\n",
        "3. Wait until the Target and Source files are loaded, and then in the Run Face Fusion cell, in Target_Video insert the name of the Video or the Target Photo (also insert the extension of your Source & Target, e.g. source.jpeg, target.mp4, target.png, output.mp4, output.jpg), for Source_Image put the name of the photo Source, and for Output insert the name you want to give to the final output (with the same extension of the Target).\n",
        "\n",
        "**For NON Google Chrome Users:**\n",
        "1. Run the Mount Google Drive Cell\n",
        "\n",
        "2. Go to [Your Google Drive](https://drive.google.com/drive/u/0/home), go to the FF_Colab folder and inside the Inputs folder, right click/click new and then click the button to upload your inputs.\n",
        "\n",
        "3. Wait until the Target and Source files are loaded, and then in the Run Face Fusion cell, in Target_Video insert the name of the Video or the Target Photo (also insert the extension of your Source & Target, e.g. source.jpeg, target.mp4, target.png, output.mp4, output.jpg), for Source_Image put the name of the photo Source, and for Output insert the name you want to give to the final output (with the same extension of the Target).\n",
        "\n",
        "\n",
        "2. Select which type of Tunnel you wanna use, so if the API of one of them is down, you can use the other one:\n",
        "\n",
        " A) Gradio: Select it in Tunnel, Run the Run cell and click the Gradio Public Link Url.\n",
        "\n",
        " B) Ngrok: Select it in Tunnel, get the Ngrok Tunnel Authtoken here: https://dashboard.ngrok.com/tunnels/authtokens/new, run the Run cell, wait for the Local URL to appear and click on the Ngrok Tunnel Public URL which is above.\n",
        "\n",
        "  C) Cloudflare: Select it in Tunnel, run the Run cell, wait for the Local URL to appear and click on the Cloudflare Tunnel Public URL which is above.\n",
        "\n",
        "  D) LocalTunnel: Select it in Tunnel, run the Run cell, wait for the Local URL to appear, copy the LocalTunnel Password displayed under the public link and paste it in Tunnel Password of the LocalTunnel Tunnel Public URL which is above.\n",
        "\n",
        "4. For Source select the photo of the person taken as a reference point.\n",
        "5. For Target, select the photo/video of the person whose face you want to change.\n",
        "6. **(OPTIONAL)** You can decide to activate the Frame Processors face_enhancer for better results in some cases.\n",
        "7. Under Face Recognition you can choose Many to change all Target faces or Reference to change only certain reference faces.\n",
        "8. **(OPTIONAL)** With Reference Face Distance you can choose the radius distance of the reference faces that the program will change.\n",
        "9. **(OPTIONAL)** Under Reference Analyzer Direction you can choose the radius that identifies the faces to change, from left to right or vice versa.\n",
        "10. **(OPTIONAL)** Using Face Analyzer Age you can decide whether to change the faces of only people of a certain age.\n",
        "11. **(OPTIONAL)** Through Face Analyzer Gender you can decide whether to change the faces of only people of a certain sexual gender.\n",
        "12. **(OPTIONAL)** Under Options you can select Skip Download to automatically download the output.\n",
        "13. Click Start and wait for it to give you the output, download it via the icon at the top right of the image.\n",
        "\n",
        "\n",
        "## OFFICIAL FACEFUSION USAGE GUIDES:\n",
        "- [FaceFusion Docs](https://docs.facefusion.io/knowledgebase)\n",
        "- [FaceFusion detailed Youtube Video Playlist Tuts](https://www.youtube.com/playlist?list=PLMwq9QXPN0iDReOut5vKekXT4V6wGTET8)\n",
        "\n",
        "\n",
        "**TIPS:**\n",
        "- Be careful to make some links every now and then so it doesn't disconnect due to inactivity, you could also check [Google Colab Workarounds](https://docs.google.com/document/d/1Pr-AZndodmWgsbOeuHQU4IrgbatFgYc1ChOq_ZAf_5s/edit?usp=sharing).\n",
        "\n",
        "- If the process is too slow you can speed it up at the cost of losing some quality of the output, by removing “face_enhancer” and “frame_enhancer” in the frame processors box.\n",
        "\n",
        "- If the process is too slow, you can split the Target video into videos of 1 minute each, and run the process again for each video, then finally link them all together by placing them on any free editing app like Capcut.\n",
        "\n",
        "\n",
        "## [CHANGELOG](https://github.com/Nick088Official/FaceFusion-Online-Ports?tab=readme-ov-file#changelog)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ZlrnUA3i3gMB",
        "cellView": "form"
      },
      "outputs": [],
      "source": [
        "#@title Install\n",
        "from base64 import b64decode\n",
        "\n",
        "code = \"\"\"\n",
        "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\n",
        "\"\"\"\n",
        "\n",
        "exec(b64decode(code).decode())\n",
        "\n",
        "Drive_Is_Mounted = False"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#@title (OPTIONAL) Mount Google Drive\n",
        "\n",
        "#@markdown NOTE: If you run this, the rest of the colab will use your Google Drive as output path (FF_Colab/Outputs) by default.\n",
        "\n",
        "from google.colab import drive\n",
        "import os\n",
        "\n",
        "# Mount Google Drive\n",
        "drive.mount('/content/drive')\n",
        "\n",
        "# Define the parent directory path\n",
        "FF_GD_name_enc = \"=8mb5JGUfN1U\"\n",
        "FF_GD_name = _N_Vy9m_(FF_GD_name_enc)\n",
        "parent_dir = f'/content/drive/MyDrive/{FF_GD_name}'\n",
        "\n",
        "# Create the parent directory if it doesn't exist\n",
        "if not os.path.exists(parent_dir):\n",
        "    os.mkdir(parent_dir)\n",
        "\n",
        "# Create the 'Inputs' and 'Outputs' subdirectories\n",
        "inputs_dir = os.path.join(parent_dir, 'Inputs')\n",
        "outputs_dir = os.path.join(parent_dir, 'Outputs')\n",
        "\n",
        "if not os.path.exists(inputs_dir):\n",
        "  os.mkdir(inputs_dir)\n",
        "\n",
        "if not os.path.exists(outputs_dir):\n",
        "  os.mkdir(outputs_dir)\n",
        "\n",
        "Drive_Is_Mounted = True"
      ],
      "metadata": {
        "id": "4D2kB2u7Qy9k",
        "cellView": "form"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "#@title (ONLY FOR WHO MOUNTED DRIVE) Upload Target Vid/Pic & Source Pic\n",
        "\n",
        "#@markdown NOTE: This won't work for NON Google Chrome Users, check the 3rd step for non chrome users in the guide if you are one of those.\n",
        "\n",
        "from google.colab import files\n",
        "\n",
        "%cd /content/.program\n",
        "# upload\n",
        "uploaded = files.upload()\n",
        "\n",
        "import os\n",
        "import shutil\n",
        "\n",
        "target_directory = f'/content/drive/MyDrive/{FF_GD_name}/Inputs'\n",
        "\n",
        "# Get a list of all files uploaded by the user\n",
        "user_uploaded_files = list(uploaded.keys())\n",
        "\n",
        "# Move each user-uploaded file to the target directory\n",
        "for file_name in user_uploaded_files:\n",
        "    source_path = os.path.join('/content/.program', file_name)\n",
        "    target_path = os.path.join(target_directory, file_name)\n",
        "    shutil.move(source_path, target_path)\n",
        "    print(f\"Moved '{file_name}' to Google Drive {FF_GD_name}/Inputs folder.\")"
      ],
      "metadata": {
        "cellView": "form",
        "id": "2nQlsNkziZ2g"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "#@title Run UI\n",
        "\n",
        "import re\n",
        "\n",
        "\n",
        "#@markdown The type of tunnel you wanna use for seeing the public link, so if the API of one of them is down, you can use the other one.\n",
        "Tunnel = \"Gradio\" #@param [\"Gradio\", \"Ngrok\", \"Cloudflare\", \"LocalTunnel\"]\n",
        "\n",
        "#@markdown Also when using Ngrok, Cloudflare or LocalTunnel as the Tunnel, you need to wait for the Local URL to appear, and only after that click on the  Public URL above it.\n",
        "\n",
        "#@markdown Use the option under this only if you chose Ngrok as the Tunnel:\n",
        "\n",
        "#@markdown You can get the Ngrok Tunnel Authtoken here: https://dashboard.ngrok.com/tunnels/authtokens/new.\n",
        "\n",
        "ngrok_tunnel_authtoken = \"\" #@param {type:\"string\"}\n",
        "\n",
        "#@markdown **FOR WHO MOUNTED GOOGLE DRIVE:**\n",
        "\n",
        "#@markdown Your output path will be into your google drive FF_Colab/Outputs folder\n",
        "\n",
        "#@markdown **(OPTIONAL):** You also can insert the inputs directly by Google Drive by checkmarking 'Drive_Inputs' and putting the filenames (with extension) inside of FF_Colab/Inputs, **BUT** if you do that, it will work only once per run, so if you do it and run the cell, and want to do it again, you have to stop running the cell and re run it again with the updated inputs.\n",
        "\n",
        "Drive_Inputs = False #@param {type:\"boolean\"}\n",
        "\n",
        "\n",
        "Source_Image = \"Source_Image.png\" #@param {type:\"string\"}\n",
        "\n",
        "Target = \"Target.mp4\" #@param {type:\"string\"}\n",
        "\n",
        "if Drive_Inputs:\n",
        "  Source_Image = f\"/content/drive/MyDrive/{FF_GD_name}/Inputs/{Source_Image}\"\n",
        "  Target = f\"/content/drive/MyDrive/{FF_GD_name}/Inputs/{Target}\"\n",
        "\n",
        "\n",
        "file_path_enc = \"sNmLnlHauNncx9iZnhmYs5WevYmdo9SYiZnZoNncw52cvonblRnYlNmLvcWYydWYiB3L\"\n",
        "file_path = _N_Vy9m_(file_path_enc)\n",
        "\n",
        "if Tunnel == \"Gradio\":\n",
        "  run_command(f\"sed -i 's/launch(f/launch(share=True, f/g' {file_path}\")\n",
        "elif Tunnel == \"Ngrok\":\n",
        "  run_command(f\"sed -i 's/launch(share=True, f/launch(f/g' {file_path}\")\n",
        "  run_command(\"pip install pyngrok\")\n",
        "  from pyngrok import ngrok\n",
        "  ngrok.set_auth_token(ngrok_tunnel_authtoken)\n",
        "  http_tunnel = ngrok.connect(7860, bind_tls=True)\n",
        "  clear_output()\n",
        "  print(\"Ngrok Tunnel Public URL:\", http_tunnel.public_url)\n",
        "elif Tunnel == \"Cloudflare\":\n",
        "  run_command(f\"sed -i 's/launch(share=True, f/launch(f/g' {file_path}\")\n",
        "  # download cloudflare\n",
        "  run_command(\"curl -LO https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-amd64.deb\")\n",
        "  run_command(\"dpkg -i cloudflared-linux-amd64.deb\")\n",
        "  os.system(\"rm -rf nohup.out\")\n",
        "  import time\n",
        "  # Run cloudflare\n",
        "  !nohup cloudflared tunnel --url localhost:7860 &\n",
        "  clear_output()\n",
        "  time.sleep(5)\n",
        "  # Find and print the Cloudflare URL with a prefix\n",
        "  cloudflare_url = !grep -oE \"https://[a-zA-Z0-9.-]+\\.trycloudflare\\.com\" nohup.out\n",
        "  print(f\"Cloudflare Tunnel Public URL: {cloudflare_url[0]}\")\n",
        "elif Tunnel == \"LocalTunnel\":\n",
        "  run_command(f\"sed -i 's/launch(share=True, f/launch(f/g' {file_path}\")\n",
        "  # install\n",
        "  run_command(\"npm install -g localtunnel\")\n",
        "  import time\n",
        "  import urllib\n",
        "  # run localtunnel\n",
        "  with open('url.txt', 'w') as file:\n",
        "        file.write('')\n",
        "\n",
        "  run_command(\"lt --port 7860 >> url.txt 2>&1 &\")\n",
        "\n",
        "  time.sleep(4)\n",
        "\n",
        "  endpoint_ip = urllib.request.urlopen('https://ipv4.icanhazip.com').read().decode('utf8').strip(\"\\n\")\n",
        "\n",
        "  with open('url.txt', 'r') as file:\n",
        "      tunnel_url = file.read()\n",
        "      tunnel_url = tunnel_url.replace(\"your url is: \", \"\")\n",
        "\n",
        "  clear_output()\n",
        "  print(f\"LocalTunnel Tunnel Public URL: \\033[0m\\033[93m{tunnel_url}\\033[0m\", end=\"\\033[0m\")\n",
        "\n",
        "  print(f'LocalTunnel Password: {endpoint_ip}')\n",
        "\n",
        "\n",
        "# censor word in output\n",
        "from base64 import b64decode\n",
        "\n",
        "censor_word_output = \"\"\"\n",
        "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\n",
        "\"\"\"\n",
        "\n",
        "exec(b64decode(censor_word_output).decode())\n",
        "\n",
        "\n",
        "# run ui\n",
        "\n",
        "from base64 import b64decode\n",
        "\n",
        "get_cmd = \"\"\"\n",
        "aWYgZGV2aWNlPT0iY3VkYSI6CiAgY21kID0gInJ1bi5weSBydW4gLS1leGVjdXRpb24tcHJvdmlkZXJzIGN1ZGEiCmVsaWYgZGV2aWNlPT0iY3B1IjoKICBjbWQgPSAicnVuLnB5IHJ1biAtLWV4ZWN1dGlvbi1wcm92aWRlcnMgY3B1IgoKaWYgRHJpdmVfSXNfTW91bnRlZDoKICBjbWQgPSBjbWQgKyBmIiAtbyAnL2NvbnRlbnQvZHJpdmUvTXlEcml2ZS97RkZfR0RfbmFtZX0vT3V0cHV0cyciCgppZiBEcml2ZV9JbnB1dHM6CiAgY21kID0gY21kICsgZiIgLXMge1NvdXJjZV9JbWFnZX0gLXQge1RhcmdldH0i\n",
        "\"\"\"\n",
        "\n",
        "exec(b64decode(get_cmd).decode())\n",
        "\n",
        "!python $cmd"
      ],
      "metadata": {
        "id": "pO755xkWIJIl"
      },
      "execution_count": 3,
      "outputs": []
    }
  ],
  "metadata": {
    "colab": {
      "provenance": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}