refactor: simplify Gradio UI layout, remove custom styling, and update demo components

app.py

@@ -373,141 +373,100 @@ def toggle(choice):
         return gr.update(visible=False, value=None), gr.update(visible=True, value=None)
 
 
-theme = gr.themes.Soft(
-    primary_hue="blue",
-    secondary_hue="indigo",
-    neutral_hue="slate",
-    font=[gr.themes.GoogleFont("Inter"), "ui-sans-serif", "sans-serif"],
-)
-
-css = """
-.gradio-container {
-    max-width: 1150px !important;
-}
-.generate-btn {
-    background: linear-gradient(135deg, #2563eb 0%, #1d4ed8 100%) !important;
-    color: white !important;
-    border: none !important;
-    font-weight: 600 !important;
-    transition: all 0.2s ease-in-out !important;
-    box-shadow: 0 4px 6px -1px rgba(37, 99, 235, 0.2) !important;
-}
-.generate-btn:hover {
-    background: linear-gradient(135deg, #1d4ed8 0%, #1e40af 100%) !important;
-    transform: translateY(-1px) !important;
-    box-shadow: 0 6px 15px rgba(37, 99, 235, 0.35) !important;
-}
-"""
-
-with gr.Blocks(theme=theme, css=css) as demo:
+with gr.Blocks() as demo:
     with gr.Row():
         gr.Markdown(
             """
-            # 🎙️ DeepFilterNet Audio Studio
-            Enhance voice recordings using **DeepFilterNet2** — an ultra-fast, real-time speech enhancement model. 
-            Adjust settings below to clean up background noise and customize vocal EQ characteristics.
+            ## DeepFilterNet2 Demo\
+
+            This demo denoises audio files using DeepFilterNet. Try it with your own voice!
             """
         )
     with gr.Row():
         with gr.Column():
-            gr.Markdown("### 📂 1. Audio Source")
             radio = gr.Radio(
-                ["mic", "file"], value="file", label="Input Method"
+                ["mic", "file"], value="file", label="How would you like to upload your audio?"
             )
-            mic_input = gr.Mic(label="Record Voice", type="filepath", visible=False)
-            audio_file = gr.Audio(type="filepath", label="Upload Audio File", visible=True)
-
-            gr.Markdown("### 🔊 2. Noise Simulation (Optional)")
-            with gr.Group():
-                noise_select = gr.Dropdown(
-                    label="Add Background Noise",
+            mic_input = gr.Mic(label="Input", type="filepath", visible=False)
+            audio_file = gr.Audio(type="filepath", label="Input", visible=True)
+            inputs = [
+                radio,
+                audio_file,
+                gr.Dropdown(
+                    label="Add background noise",
                     choices=list(NOISES.keys()),
                     value="None",
-                )
-                snr_select = gr.Dropdown(
-                    label="Simulated Noise Level (SNR)",
+                ),
+                gr.Dropdown(
+                    label="Noise Level (SNR)",
                     choices=["-5", "0", "10", "20"],
                     value="10",
-                )
-
-            gr.Markdown("### ⚙️ 3. Denoising Settings")
-            with gr.Group():
-                atten_slider = gr.Slider(
+                ),
+                gr.Slider(
                     minimum=6,
                     maximum=35,
                     step=1,
                     value=15,
-                    label="Max Attenuation (dB) - lower keeps natural room tone, higher reduces more noise",
-                )
-                blend_slider = gr.Slider(
+                    label="Max Attenuation (dB) - lower is more natural/clear, higher reduces more noise",
+                ),
+                gr.Slider(
                     minimum=50,
                     maximum=100,
                     step=5,
                     value=90,
-                    label="Voice Naturalness Mix (%) - blends original signal back to preserve transients",
-                )
-
-            with gr.Accordion("🎨 4. Advanced Vocal EQ & Presence", open=False):
-                post_process_cb = gr.Checkbox(
+                    label="Voice Naturalness Mix (%) - 100% is fully denoised, 90% blends back some original voice",
+                ),
+                gr.Checkbox(
                     label="Enable Post-Processing EQ & Filters",
                     value=True,
-                )
-                low_cut_slider = gr.Slider(
+                ),
+                gr.Slider(
                     minimum=20,
                     maximum=200,
                     step=10,
                     value=80,
-                    label="Low-Cut Filter Cutoff (Hz) - removes sub-bass rumble/AC hum",
-                )
-                bass_slider = gr.Slider(
+                    label="Low-Cut Filter Cutoff (Hz) - cuts muddy AC rumble/handling noise",
+                ),
+                gr.Slider(
                     minimum=-10,
                     maximum=10,
                     step=1,
                     value=0,
-                    label="Voice Bass Boost (dB) at 150Hz - adds body/warmth to thin voices",
-                )
-                treble_slider = gr.Slider(
+                    label="Voice Bass EQ Boost (dB) at 150Hz - adds warmth",
+                ),
+                gr.Slider(
                     minimum=-10,
                     maximum=10,
                     step=1,
                     value=3,
-                    label="Voice Treble Boost (dB) at 6kHz - adds sibilance/crispness",
-                )
-
-            inputs = [
-                radio,
-                audio_file,
-                noise_select,
-                snr_select,
-                atten_slider,
-                blend_slider,
-                post_process_cb,
-                low_cut_slider,
-                bass_slider,
-                treble_slider,
+                    label="Voice Treble EQ Boost (dB) at 6kHz - adds crispness/air",
+                ),
                 mic_input,
             ]
-            btn = gr.Button("Generate Denoised Audio", elem_classes="generate-btn")
+            btn = gr.Button("Generate")
         with gr.Column():
-            gr.Markdown("### 📊 Results & Visualization")
             outputs = [
-                gr.Audio(type="filepath", label="Noisy Audio (Input)"),
-                gr.Image(label="Noisy Spectrogram"),
-                gr.Audio(type="filepath", label="Enhanced Audio (Output)"),
-                gr.Image(label="Enhanced Spectrogram"),
+                # gr.Video(type="filepath", label="Noisy audio"),
+                gr.Audio(type="filepath", label="Noisy audio"),
+                gr.Image(label="Noisy spectrogram"),
+                # gr.Video(type="filepath", label="Enhanced audio"),
+                gr.Audio(type="filepath", label="Enhanced audio"),
+                gr.Image(label="Enhanced spectrogram"),
             ]
     btn.click(fn=demo_fn, inputs=inputs, outputs=outputs, api_name='denoise')
     radio.change(toggle, radio, [mic_input, audio_file])
     gr.Examples(
-        examples=[
-            ["./samples/p232_013_clean.wav", "Kitchen", "10"],
-            ["./samples/p232_013_clean.wav", "Cafe", "10"],
-            ["./samples/p232_019_clean.wav", "Cafe", "10"],
-            ["./samples/p232_019_clean.wav", "River", "10"],
+        [
+            ["file", "./samples/p232_013_clean.wav", "Kitchen", "10", 15, 90, True, 80, 0, 3, None],
+            ["file", "./samples/p232_013_clean.wav", "Cafe", "10", 15, 90, True, 80, 0, 3, None],
+            ["file", "./samples/p232_019_clean.wav", "Cafe", "10", 15, 90, True, 80, 0, 3, None],
+            ["file", "./samples/p232_019_clean.wav", "River", "10", 15, 90, True, 80, 0, 3, None],
         ],
-        inputs=[audio_file, noise_select, snr_select],
-        label="💡 Quick Start Demo Examples (Click to load)",
-    )
+        fn=demo_fn,
+        inputs=inputs,
+        outputs=outputs,
+        cache_examples=True,
+    ),
     gr.Markdown(open("usage.md").read())
 
 cleanup_tmp()