Update app.py

app.py

@@ -132,7 +132,24 @@ async def load_model_data():
     sess_options = onnxruntime.SessionOptions()
 
     # Collect data for all models in the directory and populate model_configurations
-    onnx_models = detect_onnx_models(models_path)
+    model_names = detect_onnx_models(models_path)
+    for model_name in model_names:
+        # Load the configuration data for each model (including speaker_id_map)
+        config = load_model_configuration(model_name)
+        model_configurations[model_name] = config
+
+
+def load_model_configuration(model_name):
+    # Assuming model_name is the path to the ONNX model file, e.g., 'model.onnx'
+    config_file_path = model_name.replace('.onnx', '.json')
+
+    try:
+        with open(config_file_path, 'r') as config_file:
+            config_data = json.load(config_file)
+        return config_data
+    except FileNotFoundError:
+        # Handle the case where the configuration file does not exist
+        return None
 
         
 #@app.post("/synthesize")
@@ -149,27 +166,7 @@ async def main(
     noise_scale_w_slider: float = Form(...),
     play: bool = Form(True)
 ):
-    """Main entry point"""
-#    sys.path.append('./content/piper/src/python')
-#    models_path = "./content/piper/src/python"
-#    logging.basicConfig(level=logging.DEBUG)
-#    providers = [
-#        "CPUExecutionProvider"
-#        if use_gpu is False
-#        else ("CUDAExecutionProvider", {"cudnn_conv_algo_search": "DEFAULT"})
-#   ]
-#    sess_options = onnxruntime.SessionOptions()
- #   model = None
-#    lang = speaker
-#    onnx_models = detect_onnx_models(models_path)
-    
-  #  if len(text_input) == 0:
-     #  text_input = "1, 2, 3. This is a test. Enter some text to generate."
-#    speaker_selection = widgets.Dropdown(
- #       options=[],
-#        description=f'{lan.translate(lang, "Select speaker")}:',
-#        layout={'visibility': 'hidden'}
-#    )
+
         # Handle loading the selected model and speaker ID here
     if selected_model in onnx_models:
         config = model_configurations[selected_model]
@@ -184,43 +181,28 @@ async def main(
         document.getElementById("synthesize_button").disabled = true;
         </script>
         """
-        onnx_model = selected_model
-        model, config = load_onnx(onnx_model, sess_options, providers)
-
-
-        config["espeak"]["voice"] = speaker
-
-        config["inference"]["noise_scale"] = noise_scale_slider
-        config["inference"]["length_scale"] = speed_slider
-        config["inference"]["noise_w"] = noise_scale_w_slider
-        
-       
- #       lan.load_language(speaker)
-        
-        print("speed: ", speed_slider)
-        print("nuber of speakers = ", config["num_speakers"])
-        print("speaker", speaker)
-        print("language", config["espeak"]["voice"])
-#        rate = speed_slider.value
- #       noise_scale = noise_scale_slider.value
-#        noise_scale_w = noise_scale_w_slider.value
+if selected_model in onnx_models:
+    # The selected_model is found in the list of model file paths
+    model_name = selected_model
+    config = load_model_configuration(model_name)
+    onnx_model = model_name  # Replace with the actual key for your ONNX model file
+    if config:
+        model, _ = load_onnx(onnx_model, sess_options, providers)
+        speaker_id_map = config.get("speaker_id_map", {})
+else:
+    # The selected_model is not found in the list; handle this case as needed
+    # You can show an error message or handle it differently
+    response_html = """
+    <div id="error-message">Selected model not found.</div>
+    <script>
+    document.getElementById("synthesize_button").disabled = true;
+    </script>
+    """
         auto_play = play
         audio = inferencing(model, config, 1, text_input, speed_slider, noise_scale_slider, noise_scale_w_slider, auto_play)
         temp_dir = tempfile.mkdtemp()
 
-        # Create a temporary directory to store the audio files
-        #temp_dir = tempfile.mkdtemp()
-
-        # Export the audio to an MP3 file in the temporary directory
-       # temp_audio_file = os.path.join(temp_dir, "generated_audio.mp3")
-        
-       # Check if text_input is more than 200 characters
-      #  if len(text_input) > 100:
-       # Truncate text_input to 200 characters
-        #   text_input = text_input[:100]
 
-        # Rename the audio file based on the text input
-     #   renamed_audio_file = os.path.join(temp_dir, f"{text_input}.mp3")
         renamed_audio_file = os.path.join(temp_dir, "download.mp3")
 
         audio.export(renamed_audio_file, format="mp3")
@@ -242,60 +224,6 @@ async def main(
         document.getElementById("synthesize_button").disabled = false;
         </script>
         """
-       # os.rename(temp_audio_file, renamed_audio_file)
-
-        # Specify the path to your MP3 audio file
-   #     audio_file_path = "path/to/your/audio.mp3"
-
-    # Check if the file exists
- #       if not os.path.exists(audio_file_path):
-  #         return {"detail": "Audio file not found"}
-
-
-#        temp_audio_file = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
-#        audio.export(temp_audio_file.name, format="mp3")
-    
-    # Rename the temporary audio file based on the text input
- #       global renamed_audio_file
-#        renamed_audio_file = os.path.join(tempfile.gettempdir(), f"{text_input}.mp3")
-#        os.rename(temp_audio_file.name, renamed_audio_file)
-        
-    else:
-        voice_model_names = []
-        for current in onnx_models:
-            voice_struct = current.split("/")[5]
-            voice_model_names.append(voice_struct)
-#        if enhanced_accessibility:
-#            playaudio("selectmodel")
-#        selection = widgets.Dropdown(
-#            options=voice_model_names,
-#            description=f'{lan.translate(lang, "Select voice package")}:',
-#        )
-#        load_btn = widgets.Button(
-#            description=lan.translate(lang, "Load it!")
-#        )
-#       config = None
-#        def load_model(button):
- #       nonlocal config
-  #      global onnx_model
-  #      nonlocal model
-   #     nonlocal models_path
-   #     selected_voice = selection.value
-     #   onnx_model = f"{models_path}/{selected_voice}"
-  #      model, config = load_onnx(onnx_model, sess_options, providers)
-#            if enhanced_accessibility:
-#                playaudio("loaded")
-#            if config["num_speakers"] > 1:
-#                speaker_selection.options = config["speaker_id_map"].values()
-#                speaker_selection.layout.visibility = 'visible'
-#                if enhanced_accessibility:
-#                    playaudio("multispeaker")
-#            else:
-#                speaker_selection.layout.visibility = 'hidden'
-
- #       load_btn.on_click(load_model)
- #       display(selection, load_btn)
-#    display(speaker_selection)