Update app.py

refactor to gradio

app.py

@@ -1,60 +1,62 @@
-import streamlit as st
-import tempfile
 import os
-from TTS.config import load_config
-from TTS.utils.manage import ModelManager
+import tempfile
+import gradio as gr
 from TTS.utils.synthesizer import Synthesizer
-from TTS.utils.download import download_url
+from huggingface_hub import hf_hub_download
 
 # Define constants
-MAX_TXT_LEN = 800
 MODEL_INFO = [
-    # ["Model Name", "Model File", "Config File", "URL"]
-    # Add other models in the same format
-    ["vits-espeak-57000", "checkpoint_57000.pth", "config.json", "https://huggingface.co/mhrahmani/persian-tts-vits-0/tree/main"],
-    # ...
+    # ["Model Name", "Model File", "Config File", "Hub URL"]
+    ["vits-espeak-57000", "checkpoint_57000.pth", "config.json", "mhrahmani/persian-tts-vits-0"],
+    # Add other models similarly...
 ]
 
+# Extract model names from MODEL_INFO
+MODEL_NAMES = [info[0] for info in MODEL_INFO]
+
+MAX_TXT_LEN = 400
+TOKEN = os.environ.get('HUGGING_FACE_HUB_TOKEN')  # Replace with the environment variable containing your token, if different
+
 # Download models
-def download_models():
-    for model_name, model_file, config_file, url in MODEL_INFO:
-        directory = model_name
-        os.makedirs(directory, exist_ok=True)
-        download_url(f"{url}{model_file}", directory, str(model_file))
-        download_url(f"{url}{config_file}", directory, "config.json")
-
-# Load a model and perform TTS
-def synthesize_speech(text, model_name):
+for model_name, model_file, config_file, repo_name in MODEL_INFO:
+    os.makedirs(model_name, exist_ok=True)
+    print(f"|> Downloading: {model_name}")
+    
+    # Use hf_hub_download to download models from Hugging Face repositories
+    hf_hub_download(repo_id=repo_name, filename=model_file, cache_dir=model_name, use_auth_token=TOKEN)
+    hf_hub_download(repo_id=repo_name, filename=config_file, cache_dir=model_name, use_auth_token=TOKEN)
+
+def synthesize(text: str, model_name: str) -> str:
+    """Synthesize speech using the selected model."""
     if len(text) > MAX_TXT_LEN:
         text = text[:MAX_TXT_LEN]
-        st.warning(f"Input text was truncated to {MAX_TXT_LEN} characters.")
-
-    synthesizer = Synthesizer(f"{model_name}/best_model.pth", f"{model_name}/config.json")
+        print(f"Input text was cut off as it exceeded the {MAX_TXT_LEN} character limit.")
+    
+    synthesizer = Synthesizer(f"{model_name}/{model_file}", f"{model_name}/{config_file}")
     if synthesizer is None:
-        st.error("Model not found!")
-        return None
+        raise NameError("Model not found")
     
     wavs = synthesizer.tts(text)
+    
     with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as fp:
         synthesizer.save_wav(wavs, fp)
         return fp.name
 
-# Streamlit app
-def main():
-    st.title('persian tts playground')
-    st.markdown("""
-    Persian TTS Demo)
-    """)
-    
-    text_input = st.text_area("Enter Text to Synthesize:", "زین همرهان سست عناصر، دلم گرفت.")
-    model_name = st.selectbox("Pick a TTS Model", [info[0] for info in MODEL_INFO], index=1)
-    
-    if st.button('Synthesize'):
-        audio_file = synthesize_speech(text_input, model_name)
-        if audio_file:
-            st.audio(audio_file, format='audio/wav')
-
-# Download models and run the Streamlit app
-if __name__ == "__main__":
-    download_models()
-    main()
+# Define Gradio interface
+iface = gr.Interface(
+    fn=synthesize,
+    inputs=[
+        gr.Textbox(label="Enter Text to Synthesize:", value="زین همرهان سست عناصر، دلم گرفت."),
+        gr.Radio(label="Pick a Model", choices=MODEL_NAMES, value=MODEL_NAMES[0]),
+    ],
+    outputs=gr.Audio(label="Output", type='filepath'),
+    examples=[["زین همرهان سست عناصر، دلم گرفت.", MODEL_NAMES[0]]],
+    title='persian tts playground',
+    description="Persian text to speech model demo",  # Add the required description here.
+    article="",
+    live=False
+)
+
+# Launch the interface
+iface.launch(share=False)
+