app.py

import os
import tempfile
import zipfile
from docx import Document
from TTS.api import TTS
from pydub import AudioSegment
import gradio as gr

# Available TTS models with voice descriptions
VOICE_MODELS = {
    "Jenny (Expressive Female)": "tts_models/en/jenny/jenny",
    "LJSpeech (Standard Female)": "tts_models/en/ljspeech/vits",
    "VCTK (Multiple Speakers)": "tts_models/en/vctk/vits"
}

# Function to update speaker choices based on the selected model
def update_speaker_choices(selected_voice):
    if selected_voice == "VCTK (Multiple Speakers)":
        return ["Speaker 1", "Speaker 2", "Speaker 3"]  # Modify with actual speaker names or indices
    return ["Default Speaker"]

def docx_to_wav_zip(doc_file, selected_voice, speaker_name):
    # Load the selected TTS model
    tts = TTS(model_name=VOICE_MODELS[selected_voice], progress_bar=False, gpu=False)

    # Extract text from .docx
    document = Document(doc_file.name)
    full_text = "\n".join([para.text for para in document.paragraphs if para.text.strip()])

    # Generate temporary paths
    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmp_wav:
        wav_path = tmp_wav.name
    zip_path = wav_path.replace(".wav", ".zip")

    # Get speaker index (this part assumes speaker names are like 'Speaker 1', 'Speaker 2', etc.)
    speaker_idx = int(speaker_name.split()[-1]) - 1 if speaker_name.startswith("Speaker") else 0

    # Generate speech with the selected speaker index
    tts.tts_to_file(text=full_text, speaker_idx=speaker_idx, file_path=wav_path)

    # Convert wav to mp3 and zip the result
    sound = AudioSegment.from_wav(wav_path)
    sound.export(wav_path, format="wav")  # keeping the wav format

    # Zip the files
    with zipfile.ZipFile(zip_path, 'w') as zipf:
        zipf.write(wav_path, os.path.basename(wav_path))

    return zip_path

# Gradio interface
interface = gr.Interface(
    fn=docx_to_wav_zip,
    inputs=[
        gr.File(label="Upload .docx File"),
        gr.Dropdown(choices=list(VOICE_MODELS.keys()), label="Choose Voice", value="Jenny (Expressive Female)"),
        gr.Dropdown(choices=update_speaker_choices("VCTK (Multiple Speakers)"), label="Choose Speaker", value="Speaker 1")  # Example
    ],
    outputs=gr.File(label="Download Zip File"),
    title="Realistic Voiceover from DOCX (Multiple Voices)",
    description="Upload a .docx file, choose a realistic voice, and pick a speaker to generate a voiceover in WAV format."
)

if __name__ == "__main__":
    interface.launch()