app.py

import sys
import os

# Make sure Python can find src/f5_tts
sys.path.append(os.path.join(os.path.dirname(__file__), "src"))

import gdown
import gradio as gr
import torch
import yaml
import numpy as np
from f5_tts.infer.utils_infer import build_model_from_config, inference

# Google Drive model + vocab links
MODEL_URL = "https://drive.google.com/uc?id=1llj4Z3uzKCXL_0EIuXqIFtXYm6lraaIz"
VOCAB_URL = "https://drive.google.com/uc?id=1YNluHbc_bqhj7B1wp9by4U0-LfCPCeLL"

# Local storage paths (download at runtime)
MODEL_PATH = "checkpoints/model_1250000.safetensors"
VOCAB_PATH = "checkpoints/vocab.txt"
CONFIG_PATH = "src/f5_tts/configs/F5TTS_Base.yaml"

# Create checkpoints folder
os.makedirs("checkpoints", exist_ok=True)

# Download model + vocab if not already present
if not os.path.exists(MODEL_PATH):
    print("🔽 Downloading model...")
    gdown.download(MODEL_URL, MODEL_PATH, quiet=False)

if not os.path.exists(VOCAB_PATH):
    print("🔽 Downloading vocab...")
    gdown.download(VOCAB_URL, VOCAB_PATH, quiet=False)

# Load model config
if not os.path.exists(CONFIG_PATH):
    raise FileNotFoundError(f"Config file not found: {CONFIG_PATH}")

with open(CONFIG_PATH, "r") as f:
    config = yaml.safe_load(f)

# Build model
model = build_model_from_config(config, MODEL_PATH, VOCAB_PATH)

# Inference function for Gradio
def infer_text(text):
    wav = inference(model, text, config)
    return (22050, np.array(wav))

# Gradio UI
demo = gr.Interface(
    fn=infer_text,
    inputs="text",
    outputs="audio",
    title="F5-TTS Text-to-Speech"
)

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