app.py

import os
from transformers import pipeline
from huggingface_hub import InferenceClient
import gradio as gr

# Initialize the translation pipeline
# Specify the source and target languages for translation
translation_pipeline = pipeline("translation", model="facebook/nllb-200-distilled-600M", src_lang="zh", tgt_lang="en")

# Initialize the text-to-speech pipeline
tts_pipeline = pipeline("text-to-speech", model="facebook/mms-tts-eng")


def translate_and_speak(chinese_text):
    """
    Translates Chinese text to English and generates speech.
    """
    # Translate the text
    # Pass the input text as a list to the pipeline
    translated_text = translation_pipeline([chinese_text])[0]['translation_text']

    # Generate speech from the translated text
    # Use a try-except block to handle potential errors during speech generation
    try:
        # The transformers text-to-speech pipeline returns a dictionary
        audio_output = tts_pipeline(translated_text)
        # The audio data is in the 'audio' key and the sampling rate in 'sampling_rate'
        audio_bytes = audio_output['audio']
        sampling_rate = audio_output['sampling_rate']
        return translated_text, (sampling_rate, audio_bytes) # Gradio expects a tuple for audio

    except Exception as e:
        return translated_text, None, f"Error generating speech: {e}"


# Create the Gradio interface
iface = gr.Interface(
    fn=translate_and_speak,
    inputs=gr.Textbox(label="Enter Chinese Text"),
    outputs=[
        gr.Textbox(label="Translated English Text"),
        gr.Audio(label="Generated Speech", format="wav")
    ],
    title="Chinese to English Translation and Text-to-Speech",
    description="Translate Chinese text to English and listen to the English translation."
)

# Launch the Gradio app
if __name__ == "__main__":
    iface.launch()