app.py

import os
os.environ["COQUI_TOS_AGREED"] = "1"

import torch.serialization
torch.serialization.add_safe_globals([
    __import__("TTS.tts.configs.xtts_config").tts.configs.xtts_config.XttsConfig,
    __import__("TTS.tts.models.xtts").tts.models.xtts.XttsAudioConfig,
    __import__("TTS.tts.models.xtts").tts.models.xtts.XttsArgs,
    __import__("TTS.config.shared_configs").config.shared_configs.BaseDatasetConfig
])

import gradio as gr
import torch
import torchaudio
from TTS.api import TTS
from pydub import AudioSegment
import uuid

# Load XTTS model
model_name = "tts_models/multilingual/multi-dataset/xtts_v2"
tts = TTS(model_name=model_name, progress_bar=False, gpu=False)

# Map emotions to file paths
emotion_to_file = {
    "Neutral": "samples/neutral.wav",
    "Sad": "samples/sad.wav",
    "Happy": "samples/happy.wav",
    "Angry": "samples/angry.wav",
    "Excited": "samples/excited.wav"
}

# Voice generator
def generate_voice(text, emotion):
    speaker_audio_path = emotion_to_file.get(emotion)
    if not os.path.isfile(speaker_audio_path):
        raise FileNotFoundError(f"Speaker audio file not found: {speaker_audio_path}")

    # Generate unique filenames to avoid overwrites
    uid = uuid.uuid4().hex
    wav_path = f"output_{uid}.wav"
    mp3_path = f"output_{uid}.mp3"

    tts.tts_to_file(
        text=text,
        speaker_wav=speaker_audio_path,
        language="en",
        file_path=wav_path
    )

    # Convert to MP3
    sound = AudioSegment.from_wav(wav_path)
    sound.export(mp3_path, format="mp3")

    return mp3_path, mp3_path

# Gradio UI
with gr.Blocks() as demo:
    gr.Markdown("## 🎙️ AI Voiceover Generator with Emotion Control")

    with gr.Row():
        script_input = gr.Textbox(label="Enter Your Script", lines=5, placeholder="Type your video script here...")
        emotion_choice = gr.Dropdown(["Neutral", "Sad", "Happy", "Angry", "Excited"], label="Select Emotion", value="Neutral")

    generate_button = gr.Button("🎤 Generate Voiceover")
    audio_output = gr.Audio(label="Listen", type="filepath")
    download_link = gr.File(label="Download MP3")

    generate_button.click(fn=generate_voice, inputs=[script_input, emotion_choice], outputs=[audio_output, download_link])

demo.launch()