Enhance language selection and error handling in transcribe_audio function

app.py

@@ -12,6 +12,7 @@ os.makedirs(MODEL_DIR, exist_ok=True)
 language_options = [(f"{code}: {name[0]}", code)
                     for code, name in LANGUAGE_CODES.items()]
 language_options.sort(key=lambda x: x[0])
+language_options = [("Auto-detect", None)] + language_options
 
 MODELS = {
     "base (140M)": "base",
@@ -69,16 +70,22 @@ def ensure_model_downloaded(model_key):
 
 
 def update_regions(language):
-    if language and language in language_to_regions:
+    if not language:
+        return [], None, False
+
+    if language in language_to_regions:
         regions = language_to_regions[language]
         regions.sort(key=lambda x: x[0])
-        return regions, regions[0][1], True
+        default_value = regions[0][1] if regions else None
+        return regions, default_value, True
     return [], None, False
 
 
 @spaces.GPU
 def transcribe_audio(audio_file, model_name, language, region, predict_timestamps, padding_speech):
     try:
+        if not audio_file:
+            return "Please upload or record audio first", ""
         ensure_assets_downloaded()
 
         model_key = MODELS[model_name]
@@ -108,6 +115,7 @@ def transcribe_audio(audio_file, model_name, language, region, predict_timestamp
             f", region: {result.region}" if result.region else "")
         return output_text, detected_info
     except Exception as e:
+        print(f"Error in transcribe_audio: {str(e)}")
         return f"Error: {str(e)}", "Transcription failed"
 
 
@@ -128,21 +136,21 @@ with gr.Blocks(title="Dolphin Speech Recognition") as demo:
             with gr.Row():
                 model_dropdown = gr.Dropdown(
                     choices=list(MODELS.keys()),
-                    value=list(MODELS.keys())[1],
+                    value=list(MODELS.keys())[0],
                     label="Model Size"
                 )
 
             with gr.Row():
                 language_dropdown = gr.Dropdown(
                     choices=language_options,
-                    value=None,
-                    label="Language (Optional)",
-                    info="If not selected, the model will auto-detect language"
+                    value=language_options[0][1],
+                    label="Language",
+                    info="Default is auto-detect"
                 )
                 region_dropdown = gr.Dropdown(
                     choices=[],
                     value=None,
-                    label="Region (Optional)",
+                    label="Region",
                     visible=False
                 )
 
@@ -162,10 +170,17 @@ with gr.Blocks(title="Dolphin Speech Recognition") as demo:
             output_text = gr.Textbox(label="Transcription", lines=10)
             language_info = gr.Textbox(label="Detected Language", lines=1)
 
+    def on_language_change(language):
+        regions, default_value, is_visible = update_regions(language)
+        return {
+            region_dropdown: gr.update(
+                choices=regions, value=default_value, visible=is_visible)
+        }
+
     language_dropdown.change(
-        fn=update_regions,
+        fn=on_language_change,
         inputs=[language_dropdown],
-        outputs=[region_dropdown, region_dropdown, region_dropdown]
+        outputs=[region_dropdown]
     )
 
     transcribe_button.click(