app.py

import gradio as gr
from TTS.api import TTS
import torch
import os

css = """
#warning {background-color: #FFCCCB !important}
.feedback label textarea {height: auto !important;
                    font-size: 22px !important;
                    font-weight: 800 !important;
                    text-align: center !important;
                    color: #801313 !important;
                    padding: 0px !important}
#alert {background-color: #fff !important}
"""

# ✅ Check CPU/GPU availability before loading models
device = "cuda" if torch.cuda.is_available() else "cpu"

# ✅ Explicitly define `gpu` settings
tts = TTS(model_name="tts_models/multilingual/multi-dataset/your_tts", progress_bar=False, gpu=(device=="cuda"))
zh_tts = TTS(model_name="tts_models/zh-CN/baker/tacotron2-DDC-GST", progress_bar=False, gpu=(device=="cuda"))
de_tts = TTS(model_name="tts_models/de/thorsten/vits", gpu=(device=="cuda"))
es_tts = TTS(model_name="tts_models/es/mai/tacotron2-DDC", progress_bar=False, gpu=(device=="cuda"))

# ✅ Ensure correct weight loading
tts.to(device)
zh_tts.to(device)
de_tts.to(device)
es_tts.to(device)

def text_to_speech(text: str, speaker_wav: str, speaker_wav_file: str) -> str:
    if len(text) > 0:
        return change_aud(text, speaker_wav, speaker_wav_file)
    else:
        return "Error: No text provided."

def change_aud(text: str, speaker_wav: str, speaker_wav_file: str) -> str:
    if speaker_wav_file and not speaker_wav:
        speaker_wav = speaker_wav_file
    
    file_path = "output.wav"

    try:
        if speaker_wav.endswith(".mp3"):
            return "Error: MP3 format not supported. Convert to WAV."

        if speaker_wav is not None:
            tts.tts_to_file(text, speaker_wav=speaker_wav, language="en", file_path=file_path)
        else:
            tts.tts_to_file(text, speaker=tts.speakers[0], language="en", file_path=file_path)

        # ✅ Debugging print statement to confirm output generation
        if os.path.exists(file_path):
            print(f"Generated file path: {file_path}, Size: {os.path.getsize(file_path)} bytes")
        else:
            print("Error: Output file was not created.")
        
        return file_path
    except Exception as e:
        return f"Error generating cloned voice: {str(e)}"

def show_error(text: str):
    return gr.update(visible=(text == ""), elem_id="warning", elem_classes="feedback")

title = "Voice-Cloning-Demo"

def toggle(choice: str):
    return (
        gr.update(visible=(choice == "mic"), value=None),
        gr.update(visible=(choice != "mic"), value=None)
    )

def change_color(text_input: str):
    return gr.update(elem_id="warning" if len(text_input) == 0 else "alert", autofocus=(len(text_input) == 0))

def clear_color(text_input: str, radio: str, error_box: str):
    return gr.update(elem_id="alert"), gr.update(value="mic"), gr.update(visible=False)

with gr.Blocks(css="footer {visibility: hidden}") as demo:
    with gr.Row():
        with gr.Column():
            text_input = gr.Textbox(label="Input the text", value="", max_lines=4, lines=4)
            radio = gr.Radio(["mic", "file"], value="mic", label="How would you like to upload your audio?")
            audio_input_mic = gr.Audio(label="Voice to clone", sources="microphone", type="filepath", visible=True)
            audio_input_file = gr.Audio(label="Voice to clone", type="filepath", visible=False)

            with gr.Row():
                with gr.Column():
                    btn_clear = gr.ClearButton([text_input, radio, audio_input_file])
                with gr.Column():
                    btn = gr.Button("Generate", variant="primary")
        with gr.Column():
            audio_output = gr.Audio(label="Output", visible=True, autoplay=True, show_share_button=False)
            error_box = gr.Textbox(label="WARNING", value="Input box cannot be blank!!", visible=False, container=True)
            
    btn_clear.add(audio_output)
    btn.click(text_to_speech, inputs=[text_input, audio_input_mic, audio_input_file], outputs=audio_output)
    btn.click(show_error, text_input, [error_box, audio_output])
    radio.change(toggle, radio, [audio_input_mic, audio_input_file])
    btn_clear.click(clear_color, [text_input, radio, error_box], [text_input, radio, error_box])
    btn.click(change_color, text_input, text_input)

demo.launch()