app.py

import os
import requests
import gradio as gr
import moviepy.editor as mp
import torch
import assemblyai as aai
from xtts_v2 import model.pth

# Import specific model components
from TTS.tts.configs.xtts_config import XttsConfig
from TTS.tts.models.xtts import Xtts

# Define paths for model and configuration files
# model_path = "/xtts_v2"
#config_path = os.path.join(model_path, "/config.json")
# checkpoint_path = model_path

# Initialize and load the XTTS model
config = XttsConfig("/xtts_v2/config.json")
model = Xtts.init_from_config(config)
model.load_checkpoint(config, checkpoint_dir="/xtts_v2", eval=True)
model.cuda()  # Move model to GPU if available

def synthesize_text(text, speaker_wav, language):
    try:
        outputs = model.synthesize(
            text,
            config,
            speaker_wav=speaker_wav,
            gpt_cond_len=3,
            language=language
        )
        return outputs
    except Exception as e:
        print(f"Error during synthesis: {e}")
        raise

# Translation class
class Translation:
    def _init_(self, video_path, original_language, target_language):
        self.video_path = video_path
        self.original_language = original_language
        self.target_language = target_language

    def org_language_parameters(self, original_language):
        language_codes = {'English': 'en', 'German': 'de', 'Italian': 'it', 'Spanish': 'es'}
        self.lan_code = language_codes.get(original_language, '')

    def target_language_parameters(self, target_language):
        language_codes = {'English': 'en', 'German': 'de', 'Italian': 'it', 'Spanish': 'es'}
        self.tran_code = language_codes.get(target_language, '')

    def extract_audio(self):
        video = mp.VideoFileClip(self.video_path)
        audio = video.audio
        audio_path = "output_audio.wav"
        audio.write_audiofile(audio_path)
        return audio_path

    def transcribe_audio(self, audio_path):
        aai.settings.api_key = os.getenv("ASSEMBLYAI_API_KEY")
        config = aai.TranscriptionConfig(language_code=self.lan_code)
        transcriber = aai.Transcriber(config=config)
        transcript = transcriber.transcribe(audio_path)
        return transcript.text

    def translate_text(self, transcript_text):
        base_url = "https://api.cognitive.microsofttranslator.com/translate"
        headers = {
            "Ocp-Apim-Subscription-Key": os.getenv("MICROSOFT_TRANSLATOR_API_KEY"),
            "Content-Type": "application/json",
            "Ocp-Apim-Subscription-Region": "southeastasia"
        }
        params = {"api-version": "3.0", "from": self.lan_code, "to": self.tran_code}
        body = [{"text": transcript_text}]
        response = requests.post(base_url, headers=headers, params=params, json=body)
        translation = response.json()[0]["translations"][0]["text"]
        return translation

    def generate_audio(self, translated_text):
        try:
            synthesized_audio = synthesize_text(
                translated_text,
                speaker_wav='output_audio.wav',
                language=self.tran_code
            )
            with open("output_synth.wav", "wb") as f:
                f.write(synthesized_audio)
            return "output_synth.wav"
        except Exception as e:
            print(f"Error generating audio: {e}")
            raise

    def translate_video(self):
        audio_path = self.extract_audio()
        self.org_language_parameters(self.original_language)
        self.target_language_parameters(self.target_language)
        transcript_text = self.transcribe_audio(audio_path)
        translated_text = self.translate_text(transcript_text)
        translated_audio_path = self.generate_audio(translated_text)

        # Run Wav2Lip inference
        os.system(f"python inference.py --checkpoint_path 'checkpoints/wav2lip_gan.pth' --face {self.video_path} --audio {translated_audio_path} --outfile 'output_video.mp4'")
        return 'output_video.mp4'


# Gradio Interface
def app(video_path, original_language, target_language):
    translator = Translation(video_path, original_language, target_language)
    video_file = translator.translate_video()
    return video_file

interface = gr.Interface(
    fn=app,
    inputs=[
        gr.Video(label="Video Path"),
        gr.Dropdown(["English", "German", "Italian", "Spanish"], label="Original Language"),
        gr.Dropdown(["English", "German", "Italian", "Spanish"], label="Targeted Language"),
    ],
    outputs=gr.Video(label="Translated Video")
)

interface.launch(share=True)  # Optional: Set share=True to create a public link