Update app.py

app.py

@@ -1,6 +1,7 @@
 # app.py
 # simple gradio space for Persian TTS using kamtera/persian-tts-female-vits (coqui tts)
-# all ui messages and comments are in English
+# loads model by first downloading the HuggingFace repo to a local folder,
+# then passes the local path to TTS to avoid Coqui's "model_name parsing" error.
 
 import os
 import tempfile
@@ -8,42 +9,50 @@ from hazm import Normalizer
 from TTS.api import TTS
 import gradio as gr
 
+# add huggingface_hub to requirements and import here
+from huggingface_hub import snapshot_download
+
 # -------------------------
 # configuration
-MODEL_ID = "Kamtera/persian-tts-female-vits"
-HF_TOKEN = os.environ.get("HF_TOKEN", None)  # optional token for private models
-MAX_INPUT_LENGTH = 1200  # safety limit for long text
+HF_REPO_ID = "Kamtera/persian-tts-female-vits"  # huggingface repo id
+HF_TOKEN = os.environ.get("HF_TOKEN", None)     # optional token for private models
+MAX_INPUT_LENGTH = 1200                          # safety limit for long text
 # -------------------------
 
 normalizer = Normalizer()
 
-# load Coqui TTS model
-print("loading tts model:", MODEL_ID)
-if HF_TOKEN:
-    tts = TTS(model_name=MODEL_ID, progress_bar=False, gpu=False, use_auth_token=HF_TOKEN)
-else:
-    tts = TTS(model_name=MODEL_ID, progress_bar=False, gpu=False)
+# download the HuggingFace repo to a local folder (cached by HF Hub)
+print("downloading model repo from huggingface:", HF_REPO_ID)
+try:
+    local_model_dir = snapshot_download(repo_id=HF_REPO_ID, use_auth_token=HF_TOKEN)
+    print("model downloaded to:", local_model_dir)
+except Exception as e:
+    print("error while downloading model repo:", e)
+    local_model_dir = None
+
+if local_model_dir is None:
+    raise RuntimeError("failed to download model repo. set HF_TOKEN if repo is private or check repo id.")
+
+# now load model from local dir (coqui expects either a coqui id or a local path)
+print("loading tts model from local folder:", local_model_dir)
+tts = TTS(model_name=local_model_dir, progress_bar=False, gpu=False)
 
 def synthesize(text: str):
     """
     text: Persian text input
-    returns: path to the generated wav file
+    returns: tuple(output_path_or_none, status_message)
     """
     if not text or not text.strip():
         return None, "please enter some text."
 
-    # limit input length to avoid high latency
     if len(text) > MAX_INPUT_LENGTH:
         text = text[:MAX_INPUT_LENGTH] + "."
 
-    # normalize persian text
     text = normalizer.normalize(text)
 
-    # create a temporary output file
     out_fd, out_path = tempfile.mkstemp(suffix=".wav")
     os.close(out_fd)
 
-    # generate audio
     try:
         tts.tts_to_file(text=text, file_path=out_path)
     except Exception as e: