Update app.py

app.py

@@ -41,7 +41,7 @@ def style_transfer(prompt, negative_prompt, audio_input):
     spec = spectro_from_wav(audio_input)
     print(spec)
     # Open the image
-    im = Image.open(spec)
+    im = image_from_spectrogram(spectrogram)
    
     new_spectro = pipe2(prompt=prompt, image=im, strength=0.5, guidance_scale=7).images
     wav = wav_bytes_from_spectrogram_image(new_spectro[0])
@@ -49,6 +49,32 @@ def style_transfer(prompt, negative_prompt, audio_input):
         f.write(wav[0].getbuffer())
     return new_spectro, 'output.wav', gr.update(visible=True), gr.update(visible=True), gr.update(visible=True)
 
+def image_from_spectrogram(
+    spectrogram: np.ndarray, max_volume: float = 50, power_for_image: float = 0.25
+) -> Image.Image:
+    """
+    Compute a spectrogram image from a spectrogram magnitude array.
+    """
+    # Apply the power curve
+    data = np.power(spectrogram, power_for_image)
+
+    # Rescale to 0-255
+    data = data * 255 / max_volume
+
+    # Invert
+    data = 255 - data
+
+    # Convert to a PIL image
+    image = Image.fromarray(data.astype(np.uint8))
+
+    # Flip Y
+    image = image.transpose(Image.FLIP_TOP_BOTTOM)
+
+    # Convert to RGB
+    image = image.convert("RGB")
+
+    return image
+
 title = """
     <div style="text-align: center; max-width: 500px; margin: 0 auto;">
         <div