import gradio as gr
import torch
import whisper
import requests
from pytube import YouTube

### ————————————————————————————————————————

title="Transcript PDF"

### ————————————————————————————————————————

whisper_model = whisper.load_model("medium")

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

def transcribe(audio):
    
    print("""
    —
    Sending audio to Whisper ...
    —
    """)
    
    audio = whisper.load_audio(audio)
    audio = whisper.pad_or_trim(audio)
    
    mel = whisper.log_mel_spectrogram(audio).to(whisper_model.device)
    
    _, probs = whisper_model.detect_language(mel)
    
    transcript_options = whisper.DecodingOptions(task="transcribe", fp16 = False)
    translate_options = whisper.DecodingOptions(task="translate", fp16 = False)
    
    transcription = whisper.decode(whisper_model, mel, transcript_options)
    translation = whisper.decode(whisper_model, mel, translate_options)
    
    print("Language Spoken: " + transcription.language)
    print("Transcript: " + transcription.text)  
    print("Translated: " + translation.text)

 
    return transcription.text

def transcribe_upload(audio):
    return transcribe(audio)

def transcribe_yt(link):
    yt = YouTube(link)
    path = yt.streams.filter(only_audio=True)[0].download(filename="audio.mp3")
    return transcribe(path)

with gr.Blocks(css = css) as demo:
    gr.Markdown("""
                ## Multi-lingual Transcript Generator
            """)
    gr.HTML('''
     <p style="margin-bottom: 10px">
                Save Transcripts of videos as PDF with the help of Whisper, which is a general-purpose speech recognition model released by OpenAI that can perform multilingual speech recognition as well as speech translation and language identification. 
              </p>
              ''')
    with gr.Column():
            #gr.Markdown(""" ### Record audio """)
        with gr.Tab("Youtube Link"):
            yt_input = gr.Textbox(label = 'Youtube Link')
            transcribe_audio_yt = gr.Button('Transcribe')
        
        with gr.Tab("Upload Podcast as File"):
            audio_input_u = gr.Audio(label = 'Upload Audio',source="upload",type="filepath")
            transcribe_audio_u = gr.Button('Transcribe')

        with gr.Row():
            transcript_output = gr.Textbox(label="Transcription in the language spoken", lines = 20)
            summary_output = gr.Textbox(label = "English Summary", lines = 10)
    
    transcribe_audio_yt.click(transcribe_yt, inputs = yt_input, outputs = [transcript_output, summary_output])
    transcribe_audio_u.click(transcribe_upload, inputs = audio_input_u, outputs = [transcript_output,summary_output])   
    gr.HTML('''
        <div class="footer">
                    <p>Whisper Model by <a href="https://github.com/openai/whisper" style="text-decoration: underline;" target="_blank">OpenAI</a>
                    </p>
        </div>
        ''')
        
demo.queue()    
demo.launch()