added examples

.gitattributes

@@ -4,3 +4,4 @@ models/diffusion_pitch/last.ckpt filter=lfs diff=lfs merge=lfs -text
 models/pitch_to_audio/last.ckpt filter=lfs diff=lfs merge=lfs -text
 models/diffusion_pitch/qt.joblib filter=lfs diff=lfs merge=lfs -text
 models/pitch_to_audio/qt.joblib filter=lfs diff=lfs merge=lfs -text
+examples/** filter=lfs diff=lfs merge=lfs -text

app.py

@@ -195,14 +195,46 @@ def set_guide_and_generate(audio):
     return audio, user_input_plot, pitch
 
 with gr.Blocks() as demo:
+    with gr.Column():
+        gr.Markdown("""
+                    # GaMaDHaNi: HIERARCHICAL GENERATIVE MODELING OF MELODIC VOCAL CONTOURS IN HINDUSTANI CLASSICAL MUSIC
+                    :book: Read more about the project [here](https://arxiv.org/pdf/2408.12658) <br>
+                    :samples: Listen to the samples [here](https://snnithya.github.io/gamadhani-samples) <br>
+                    # """)
+        gr.Markdown("""
+                    ## Instructions
+                    In this demo you can interact with the model in two ways: 
+                    1. **Call and response**: The model will try to continue the idea that you input. This is similar to `primed generation' discussed in the paper.
+                    2. **Melodic reinterpretation**: Akin to the idea of `coarse pitch conditioning' presented in the paper, you can input a pitch contour and the model will generate audio that is similar to but not exactly the same. <br><br>
+                    **Upload an audio file or record your voice to get started!**
+                    """)
+        gr.Markdown("""
+                    This is still a work in progress, so please feel free to share any weird or interesting examples, we would love to hear them! Contact us at [snnithya.mit.edu](mailto:snnithya.mit.edu).
+                    """)
+
     with gr.Row():
         with gr.Column():    
             audio = gr.Audio(label="Input")
             sbmt = gr.Button()
-            user_input = gr.Plot(label="User Input")
+            with gr.Accordion("View Pitch Plot"):
+                user_input = gr.Plot(label="User Input")
         with gr.Column():
             generated_audio = gr.Audio(label="Generated Audio")
-            generated_pitch = gr.Plot(label="Generated Pitch")
+            with gr.Accordion("View Pitch Plot"):
+                generated_pitch = gr.Plot(label="Generated Pitch")
+    example_description = gr.Textbox(label="Example Description", interactive=False)
+    examples = gr.Examples(
+        examples=[
+            ["examples/ex1.wav"],
+            ["examples/ex2.wav"],
+            ["examples/ex3.wav"],
+            ["examples/ex4.wav"],
+            ["examples/ex5.wav"]
+            # Add more examples as needed
+        ],
+        inputs=audio
+    )
+
     sbmt.click(set_guide_and_generate, inputs=[audio], outputs=[generated_audio, user_input, generated_pitch])
 
 def main(argv):

examples/ex1-hf.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30ae54c7e54036729d69fe70bfb294d1b80e445ad917df40bb293dfd88e7b21a
+size 307406

examples/ex2-hf.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9705358da218e84a8879b31ff30ee4a692706b6a5b6f76a9fdb21075300e3e43
+size 265080

examples/ex3-hf.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e816747fc71742ac1a4897ff7eebc3a7214c5d23a3c1b291f6d96e0630b82a9d
+size 159868

examples/ex4-hf.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0178c8ca78f6ae4bcd312f3f5c005dabfd1cb1f1520802872a4cd0507fa5878d
+size 291182

examples/ex5-hf.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e65935858bd0f9a8665d90929d69b8f2fcd6a5f52f079f41542c57426c8b4ace
+size 128950