app.py

import os
from huggingface_hub import notebook_login
import gradio as gr
from transformers import pipeline
from deep_translator import GoogleTranslator
from gtts import gTTS
import shutil
from pytube import YouTube


pipe = pipeline(
  "automatic-speech-recognition",
  model="nelanbu/ID2223_Lab2_Whisper",
  chunk_length_s=30,
  batch_size=16,
  stride_length_s=(4, 2),
  device_map="auto")  

def transcribe(audio_input, link_input, lang):
    try:

        if link_input:
          # video=YouTube(link_input).streams.filter(only_audio=True).all()
          # audio=video[0].download()
          audio_streams = YouTube(link_input).streams.filter(only_audio=True).order_by('abr').desc()
          best_audio = audio_streams.first()
          audio=best_audio.download()


        elif audio_input:
          audio = audio_input
          # input_path = "input.mp3"
          # shutil.copyfile(audio, input_path)
          # print(f"Input audio is saved to {input_path}")

        result = pipe(audio)
        text = result['text']
        print(f"Transcribed text: {text}")

        if lang == 'english':
          target_lang = 'en'
        elif lang == 'swedish':
          target_lang = 'sv'
        elif lang == 'italian':
          target_lang = 'it'
        elif lang == 'german':
          target_lang = 'de'
        elif lang == 'spanish':
          target_lang = 'es'
        elif lang == 'french':
          target_lang = 'fr'
        elif lang == 'danish':
          target_lang = 'da'
        elif lang == 'russian':
          target_lang = 'ru'
        elif lang == 'japanese':
          target_lang = 'ja'
        elif lang == 'chinese (simplified)':
          target_lang = 'zh-cn'

        translator = GoogleTranslator(source='auto', target=target_lang)
        translated_text = translator.translate(text)
        print(f"Translated text: {translated_text}")
        # Convert translated text to speech
        tts = gTTS(text=translated_text, tld='com', slow=False, lang=target_lang)
        output_audio_path = "test2.mp3"
        tts.save(output_audio_path)
        print(f"Saved TTS audio to {output_audio_path}")

        return text, translated_text, output_audio_path
    except Exception as e:
        print(f"An error occurred: {e}")
        raise

demo = gr.Interface(
    fn=transcribe,
    inputs=[gr.Audio(type="filepath", label="Translate from microphone/MP3 input"),
            gr.Text(max_lines=1, label="Translate from YouTube URL"),
            gr.Dropdown(
            ["english", "swedish", "italian", "german", "spanish", "french", "danish", "russian", "japanese", "chinese"], label="Language", info="pick the language you want to translate your auido"),
            ],
    outputs=[gr.Text(label="transcribed text"), gr.Text(label="translated text"), gr.Audio(type="filepath", label="translated audio")],
    title="Turkish Audio Translator",
    description="You can upload YouTube link of a video in Turkish or use the microphone to record your voice or upload an MP3 file to translate Turkish audio to other languages."
)

if __name__ == "__main__":
    demo.launch(debug=True)