U

app.py

@@ -6,6 +6,7 @@ import soundfile as sf
 import spaces
 import torch
 import torchaudio
+from gradio.themes.utils import colors
 from sv import process_audio
 
 
@@ -34,15 +35,21 @@ def model_inference(input_wav, language):
 
 
 def launch():
-    with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    # Create a custom orange theme
+    orange_theme = gr.themes.Default().set(
+        primary_hue=colors.orange,
+        secondary_hue=colors.stone,
+    )
+
+    with gr.Blocks(theme=orange_theme) as demo:
         gr.Markdown("# Cantonese Call Transcriber")
         gr.Markdown("## Try an example:")
 
         # Define components
-        audio_inputs = gr.Audio(label="Input")
-        text_outputs = gr.Textbox(lines=10, label="Output")
+        audio_inputs = gr.Audio(label="Input", visible=False)
+        text_outputs = gr.Textbox(lines=10, label="Output", visible=False)
 
-        # Place the Examples component above the input and output
+        # Place the Examples component first
         gr.Examples(
             examples=[["example/scb.mp3"]],
             inputs=[audio_inputs],
@@ -51,15 +58,21 @@ def launch():
             examples_per_page=1,
         )
 
+        # Main interface
         with gr.Row():
             with gr.Column(scale=2):
-                # Audio input is already defined, just reference it here
-                audio_inputs
+                gr.Audio(label="Input")
                 fn_button = gr.Button("Process Audio", variant="primary")
 
             with gr.Column(scale=3):
-                # Text output is already defined, just reference it here
-                text_outputs
+                gr.Textbox(lines=10, label="Output")
+
+        # Set up event handler
+        fn_button.click(
+            fn=lambda x: model_inference(x, "yue"),
+            inputs=[audio_inputs],
+            outputs=[text_outputs],
+        )
 
     demo.launch()