Update app.py

app.py

@@ -28,18 +28,29 @@ def predict(prompt, negative_prompt, audio_input, duration):
     # return style_transfer(prompt, negative_prompt, audio_input)
 
 def classic(prompt, negative_prompt, duration):
-    pipe.safety_checker = dummy_checker
-    spec = pipe(prompt, negative_prompt=negative_prompt, height=512, width=512).images[0]
+    pipe2.safety_checker = dummy_checker
+    url = "https://huggingface.co/spaces/gfartenstein/text2fart/resolve/main/rootfart-1.jpg"
+    response = requests.get(url)
+    im = Image.open(BytesIO(response.content)).convert("RGB")
+    # spec = pipe(prompt, negative_prompt=negative_prompt, height=512, width=512).images[0]
+    spec = pipe2(prompt=prompt, negative_prompt=negative_prompt, image=im, strength=0.5, guidance_scale=7).images
     print(spec)
     wav = wav_bytes_from_spectrogram_image(spec)
     with open("output.wav", "wb") as f:
         f.write(wav[0].getbuffer())
     return spec, 'output.wav', gr.update(visible=True), gr.update(visible=True), gr.update(visible=True)
 
-def style_transfer(prompt, negative_prompt, audio_input):
+# def style_transfer(prompt, negative_prompt, audio_input):
+    # pipe.safety_checker = dummy_checker
+    # url = "https://huggingface.co/spaces/gfartenstein/text2fart/resolve/main/rootfart-1.jpg"
+    # response = requests.get(url)
+    # init_image = Image.open(BytesIO(response.content)).convert("RGB")
+    # images = pipe(prompt=prompt, image=init_image, strength=0.75, guidance_scale=7.5).images
+
+    
     # spec = spectro_from_wav(audio_input)
     # Open the image
-    im = Image.open('rootfart-1.jpg')
+    # im = Image.open('rootfart-1.jpg')
     # im = Image.open(spec)
     
     
@@ -47,37 +58,37 @@ def style_transfer(prompt, negative_prompt, audio_input):
     # im = image_from_spectrogram(im, 1)
    
     
-    new_spectro = pipe2(prompt=prompt, image=im, strength=0.5, guidance_scale=7).images
-    wav = wav_bytes_from_spectrogram_image(new_spectro[0])
-    with open("output.wav", "wb") as f:
-        f.write(wav[0].getbuffer())
-    return new_spectro[0], 'output.wav', gr.update(visible=True), gr.update(visible=True), gr.update(visible=True)
+    # new_spectro = pipe2(prompt=prompt, image=im, strength=0.5, guidance_scale=7).images
+    # wav = wav_bytes_from_spectrogram_image(new_spectro[0])
+    # with open("output.wav", "wb") as f:
+    #     f.write(wav[0].getbuffer())
+    # return new_spectro[0], '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)
+# 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
+#     # Rescale to 0-255
+#     data = data * 255 / max_volume
 
-    # Invert
-    data = 255 - data
+#     # Invert
+#     data = 255 - data
 
-    # Convert to a PIL image
-    image = Image.fromarray(data.astype(np.uint8))
+#     # Convert to a PIL image
+#     image = Image.fromarray(data.astype(np.uint8))
 
-    # Flip Y
-    image = image.transpose(Image.FLIP_TOP_BOTTOM)
+#     # Flip Y
+#     image = image.transpose(Image.FLIP_TOP_BOTTOM)
 
-    # Convert to RGB
-    image = image.convert("RGB")
+#     # Convert to RGB
+#     image = image.convert("RGB")
 
-    return image
+#     return image
 
 title = """
     <div style="text-align: center; max-width: 500px; margin: 0 auto;">