app.py

import os
import tempfile
import threading
import gradio as gr
from TTS.api import TTS

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

MODEL_PATH = "models/xtts"
CONFIG_PATH = "models/xtts/config.json"
VOCAB_PATH = "models/xtts/vocab.json"
SPEAKERS_PATH = "models/xtts/speakers_xtts.pth"

for p in [MODEL_PATH, CONFIG_PATH, VOCAB_PATH, SPEAKERS_PATH]:
    if not os.path.exists(p):
        raise RuntimeError(f"Missing file: {p}")

tts = None
model_ready = False


def load_model():
    global tts, model_ready
    print("🔁 Loading XTTS model once...")

    tts = TTS(
        model_path=MODEL_PATH,
        config_path=CONFIG_PATH,
        progress_bar=False,
        gpu=False
    )

    model_ready = True
    print("✅ XTTS model loaded!")


threading.Thread(target=load_model, daemon=True).start()


def tts_api(text, speaker_wav):

    if not model_ready:
        raise gr.Error("Model loading... wait 30 seconds")

    if not text or speaker_wav is None:
        raise gr.Error("Provide text and speaker audio")

    out_path = os.path.join(tempfile.gettempdir(), "out.wav")

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

    return out_path


demo = gr.Interface(
    fn=tts_api,
    inputs=[
        gr.Textbox(label="Text"),
        gr.Audio(type="filepath", label="Speaker WAV"),
    ],
    outputs=gr.Audio(type="filepath"),
    api_name="tts"
)

demo.queue()
demo.launch()