app.py

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

# Fixed model (YourTTS in English)
YOURTTS_MODEL = "tts_models/multilingual/multi-dataset/your_tts"

# Fixed speaker file (pre-cloned voice)
FIXED_SPEAKER_PATH = "speakers/voice2.wav"

# Initialize model once
tts = TTS(YOURTTS_MODEL, gpu=False)

def synthesize(text):
    output_path = "output.wav"
    start_time = time.time()

    if not os.path.exists(FIXED_SPEAKER_PATH):
        return None, {"error": f"❌ Speaker file not found: {FIXED_SPEAKER_PATH}"}

    try:
        tts.tts_to_file(
            text=text,
            speaker_wav=FIXED_SPEAKER_PATH,
            file_path=output_path,
            language="en"
        )
    except Exception as e:
        return None, {"error": str(e)}

    total_time = time.time() - start_time
    est_duration = len(text.split()) / 2.5
    rtf = round(total_time / est_duration, 3)

    return output_path, {
        "language": "English",
        "processing_time_sec": round(total_time, 3),
        "real_time_factor": rtf,
        "model_used": YOURTTS_MODEL,
        "speaker_used": FIXED_SPEAKER_PATH.split("/")[-1]
    }

# ------------------ Gradio UI ------------------
with gr.Blocks() as demo:
    gr.Markdown("## 🗣️ YourTTS Voice Cloning (English Only, Fixed Speaker)")

    input_text = gr.Textbox(
        label="Text",
        placeholder="Type something to synthesize...",
        lines=3
    )

    generate_btn = gr.Button("🔊 Generate Speech")

    output_audio = gr.Audio(label="Output Audio", type="filepath")
    metadata_json = gr.JSON(label="Meta Info (Time, Model, RTF, etc.)")

    generate_btn.click(
        fn=synthesize,
        inputs=[input_text],
        outputs=[output_audio, metadata_json]
    )

# ------------------ API Interface ------------------
api_demo = gr.Interface(
    fn=synthesize,
    inputs=[gr.Text(label="Text")],   # Only text input
    outputs=[
        gr.Audio(type="filepath", label="Generated Audio"),
        gr.JSON(label="Metadata")
    ],
    title="YourTTS Voice Cloning (English Only, Fixed Speaker)"
)

# Launch the app with both UI and API
demo.launch(server_name="0.0.0.0", server_port=7860, show_api=True)