{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "kPCLdTfJyktF"
      },
      "outputs": [],
      "source": [
        "import torch\n",
        "\n",
        "import gradio as gr\n",
        "import pytube as pt\n",
        "from transformers import pipeline\n",
        "\n",
        "asr = pipeline(\n",
        "    task=\"automatic-speech-recognition\",\n",
        "    model=\"Yasaman/whisper_fa\",\n",
        "    chunk_length_s=30,\n",
        "    device=\"cpu\",\n",
        ")\n",
        "\n",
        "summarizer = pipeline(\n",
        "    \"summarization\",\n",
        "    model=\"alireza7/PEGASUS-persian-base-PN-summary\",\n",
        ")\n",
        "\n",
        "translator = pipeline(\n",
        "    \"translation\", \n",
        "    model=\"Helsinki-NLP/opus-mt-iir-en\")\n",
        "\n",
        "def transcribe(microphone, file_upload):\n",
        "    warn_output = \"\"\n",
        "    if (microphone is not None) and (file_upload is not None):\n",
        "        warn_output = (\n",
        "            \"WARNING: You've uploaded an audio file and used the microphone. \"\n",
        "            \"The recorded file from the microphone will be used and the uploaded audio will be discarded.\\n\"\n",
        "        )\n",
        "\n",
        "    elif (microphone is None) and (file_upload is None):\n",
        "        return \"ERROR: You have to either use the microphone or upload an audio file\"\n",
        "\n",
        "    file = microphone if microphone is not None else file_upload\n",
        "\n",
        "    text = asr(file)[\"text\"]\n",
        "\n",
        "    translate = translator(text)\n",
        "    translate = translate[0][\"translation_text\"]\n",
        "\n",
        "    return warn_output + text, translate\n",
        "\n",
        "def _return_yt_html_embed(yt_url):\n",
        "    video_id = yt_url.split(\"?v=\")[-1]\n",
        "    HTML_str = (\n",
        "        f'<center> <iframe width=\"500\" height=\"320\" src=\"https://www.youtube.com/embed/{video_id}\"> </iframe>'\n",
        "        \" </center>\"\n",
        "    )\n",
        "    return HTML_str\n",
        "\n",
        "\n",
        "def yt_transcribe(yt_url):\n",
        "    yt = pt.YouTube(yt_url)\n",
        "    html_embed_str = _return_yt_html_embed(yt_url)\n",
        "    stream = yt.streams.filter(only_audio=True)[0]\n",
        "    stream.download(filename=\"audio.mp3\")\n",
        "\n",
        "    text = asr(\"audio.mp3\")[\"text\"]\n",
        "\n",
        "    summary = summarizer(text)\n",
        "    summary = summary[0][\"summary_text\"]\n",
        "      \n",
        "    translate = translator(summary)\n",
        "    translate = translate[0][\"translation_text\"]\n",
        "\n",
        "    return html_embed_str, text, summary, translate\n",
        "\n",
        "demo = gr.Blocks()\n",
        "\n",
        "mf_transcribe = gr.Interface(\n",
        "    fn=transcribe,\n",
        "    inputs=[\n",
        "        gr.inputs.Audio(source=\"microphone\", type=\"filepath\", optional=True),\n",
        "        gr.inputs.Audio(source=\"upload\", type=\"filepath\", optional=True),\n",
        "    ],\n",
        "    outputs=[\n",
        "             gr.Textbox(label=\"Transcribed text\"),\n",
        "             gr.Textbox(label=\"Translated text\"),\n",
        "    ],\n",
        "    layout=\"horizontal\",\n",
        "    theme=\"huggingface\",\n",
        "    title=\"Whisper Demo: Transcribe and Translate Persian Audio\",\n",
        "    description=(\n",
        "        \"Transcribe and Translate long-form microphone or audio inputs with the click of a button! Demo uses the the fine-tuned\"\n",
        "        f\" [Yasaman/whisper_fa](https://huggingface.co/Yasaman/whisper_fa) and 🤗 Transformers to transcribe audio files\"\n",
        "        \" of arbitrary length. It also uses another model for the translation.\"\n",
        "    ),\n",
        "    allow_flagging=\"never\",\n",
        ")\n",
        "\n",
        "yt_transcribe = gr.Interface(\n",
        "    fn=yt_transcribe,\n",
        "    inputs=[gr.inputs.Textbox(lines=1, placeholder=\"Paste the URL to a YouTube video here\", label=\"YouTube URL\")],\n",
        "    outputs=[\"html\",\n",
        "             gr.Textbox(label=\"Transcribed text\"),\n",
        "             gr.Textbox(label=\"Summarized text\"),\n",
        "             gr.Textbox(label=\"Translated text\"),\n",
        "    ],\n",
        "    layout=\"horizontal\",\n",
        "    theme=\"huggingface\",\n",
        "    title=\"Whisper Demo: Transcribe, Summarize and Translate YouTube\",\n",
        "    description=(\n",
        "        \"Transcribe, Summarize and Translate long-form YouTube videos with the click of a button! Demo uses the the fine-tuned \"\n",
        "        f\" [Yasaman/whisper_fa](https://huggingface.co/Yasaman/whisper_fa) and 🤗 Transformers to transcribe audio files of\"\n",
        "        \" arbitrary length. It also uses other two models to first summarize and then translate the text input. You can try with the following example: \" \n",
        "        f\" [Video1](https://www.youtube.com/watch?v=qtRzP3KvQZk)\"\n",
        "    ),\n",
        "    allow_flagging=\"never\",\n",
        ")\n",
        "\n",
        "with demo:\n",
        "    gr.TabbedInterface([mf_transcribe, yt_transcribe], [\"Transcribe and Translate Audio\", \"Transcribe, Summarize and Translate YouTube\"])\n",
        "\n",
        "demo.launch(enable_queue=True)"
      ]
    }
  ]
}