Update app.py

app.py

@@ -1,28 +1,32 @@
 import os
 import gradio as gr
-from TTS.api import TTS
 import torch
+from TTS.tts.configs.xtts_config import XttsConfig
+from TTS.tts.models.xtts import Xtts
 
 os.environ["COQUI_TOS_AGREED"] = "1"
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-print("Loading XTTS...")
-try:
-    # Syntax 1: Standard ID
-    tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
-except Exception as e:
-    print(f"Error 1: {e}")
-    try:
-        # Syntax 2: Explicit model_name kwarg
-        tts = TTS(model_name="tts_models/multilingual/multi-dataset/xtts_v2", progress_bar=False).to(device)
-    except Exception as e2:
-        print(f"Error 2: {e2}")
-        # Syntax 3: Short ID (sometimes works depending on mapping)
-        tts = TTS("xtts_v2").to(device)
+print("Loading XTTS from Hugging Face Hub...")
+config = XttsConfig()
+config.load_json("https://huggingface.co/coqui/XTTS-v2/raw/main/config.json")
+model = Xtts.init_from_config(config)
+model.load_checkpoint(config, checkpoint_dir="coqui/XTTS-v2", use_deepspeed=False)
+if device == "cuda":
+    model.cuda()
 
 def clone(text, lang, ref):
     out = "out.wav"
-    tts.tts_to_file(text=text, file_path=out, speaker_wav=ref, language=lang)
+    outputs = model.synthesize(
+        text,
+        config,
+        speaker_wav=ref,
+        language=lang,
+        gpt_cond_len=3,
+        temperature=0.75
+    )
+    import scipy.io.wavfile
+    scipy.io.wavfile.write(out, 24000, outputs["wav"])
     return out
 
 LANGS = ["en","es","fr","de","it","pt","pl","tr","ru","nl","cs","ar","zh-cn","ja","ko","hu"]