app.py

import gradio as gr
import whisper
from transformers import MarianMTModel, MarianTokenizer
import subprocess
import os

# Load models
def load_models():
    global whisper_model, translation_models, translation_tokenizers
    whisper_model = whisper.load_model("base")  # Whisper model
    translation_models = {
        "en": MarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-en-es"),
        "es": MarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-es-en"),
        "fr": MarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-fr-en"),
        "ar": MarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-ar-en"),
    }
    translation_tokenizers = {
        lang: MarianTokenizer.from_pretrained(f"Helsinki-NLP/opus-mt-{lang}-en")
        for lang in translation_models.keys()
    }

load_models()

# Transcribe function
def transcribe_audio(file, language="en"):
    try:
        result = whisper_model.transcribe(file, language=language)
        transcription = result["text"]
        return transcription
    except Exception as e:
        return f"Error: {str(e)}"

# Translate function
def translate_text(text, source_lang, target_lang="en"):
    try:
        if source_lang not in translation_models or target_lang != "en":
            return "Unsupported language."

        tokenizer = translation_tokenizers[source_lang]
        model = translation_models[source_lang]
        inputs = tokenizer(text, return_tensors="pt", padding=True)
        translated_tokens = model.generate(**inputs)
        translated_text = tokenizer.decode(translated_tokens[0], skip_special_tokens=True)
        return translated_text
    except Exception as e:
        return f"Error: {str(e)}"

# Text-to-Speech function
def text_to_speech(text, speaker="male", speed="normal"):
    try:
        output_file = "output.wav"
        tts_command = [
            "tts",
            f"--text={text}",
            "--model_name=tts_models/en/ljspeech/tacotron2-DCA",
            f"--out_path={output_file}",
        ]
        subprocess.run(tts_command, check=True)
        return output_file
    except Exception as e:
        return f"Error: {str(e)}"

# Gradio Interface
def tts_interface(text):
    audio_file = text_to_speech(text)
    return audio_file if isinstance(audio_file, str) and os.path.exists(audio_file) else None

with gr.Blocks() as demo:
    gr.Markdown("### Audio Transcription, Translation, and TTS App")
    
    # Transcription section
    with gr.Row():
        with gr.Column():
            audio_input = gr.Audio(label="Upload Audio File", type="file")
            lang_input = gr.Dropdown(["en", "es", "fr", "ar"], label="Language", value="en")
            transcribe_btn = gr.Button("Transcribe")
        transcription_output = gr.Textbox(label="Transcription Output")

    transcribe_btn.click(transcribe_audio, inputs=[audio_input, lang_input], outputs=transcription_output)

    # Translation section
    with gr.Row():
        with gr.Column():
            text_input = gr.Textbox(label="Input Text", lines=3)
            source_lang = gr.Dropdown(["en", "es", "fr", "ar"], label="Source Language", value="en")
            translate_btn = gr.Button("Translate")
        translation_output = gr.Textbox(label="Translation Output")

    translate_btn.click(translate_text, inputs=[text_input, source_lang], outputs=translation_output)

    # TTS section
    with gr.Row():
        with gr.Column():
            tts_input = gr.Textbox(label="Text for TTS", lines=2)
            tts_btn = gr.Button("Generate Audio")
        tts_output = gr.Audio(label="Generated Audio")

    tts_btn.click(tts_interface, inputs=tts_input, outputs=tts_output)

# Launch Gradio App
demo.launch()