Create app.py

app.py

@@ -0,0 +1,56 @@
+import gradio as gr
+import os
+import tempfile
+from TTS.api import TTS
+
+print("🚀 Starting XTTS Space...")
+
+MODEL_NAME = "tts_models/multilingual/multi-dataset/xtts_v2"
+
+# 🔒 Global singleton (loaded once)
+tts = None
+
+def load_model_once():
+    global tts
+    if tts is None:
+        print("⬇️ Downloading & loading XTTS model (one-time)...")
+        tts = TTS(model_name=MODEL_NAME, gpu=False)
+        print("✅ XTTS model loaded.")
+    return tts
+
+def generate_audio(text, speaker_wav):
+    if not text or speaker_wav is None:
+        return None
+
+    tts = load_model_once()
+
+    # Output file
+    out_path = os.path.join(tempfile.gettempdir(), "output.wav")
+
+    tts.tts_to_file(
+        text=text,
+        speaker_wav=speaker_wav,
+        language="hi",
+        file_path=out_path
+    )
+
+    return out_path
+
+with gr.Blocks(title="XTTS Voice Clone (CPU)") as demo:
+    gr.Markdown("## 🎙️ XTTS Voice Cloning (CPU Space)\nUpload voice + enter text → get WAV")
+
+    with gr.Row():
+        text_in = gr.Textbox(label="Enter Text (Hindi / English)", lines=4, placeholder="नमस्ते...")
+        wav_in = gr.Audio(label="Upload Voice Sample (WAV)", type="filepath")
+
+    out_audio = gr.Audio(label="Generated Voice")
+
+    btn = gr.Button("Generate Voice")
+
+    btn.click(
+        fn=generate_audio,
+        inputs=[text_in, wav_in],
+        outputs=out_audio
+    )
+
+demo.queue(concurrency_count=1).launch()