Update app.py

app.py

@@ -12,10 +12,6 @@ from pipeline_fill_sd_xl import StableDiffusionXLFillPipeline
 from PIL import Image, ImageDraw
 import numpy as np
 
-MODELS = {
-    "RealVisXL V5.0 Lightning": "SG161222/RealVisXL_V5.0_Lightning",
-}
-
 config_file = hf_hub_download(
     "xinsir/controlnet-union-sdxl-1.0",
     filename="config_promax.json",
@@ -47,12 +43,18 @@ pipe = StableDiffusionXLFillPipeline.from_pretrained(
 
 pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
 
+def can_expand(source_width, source_height, target_width, target_height, alignment):
+    """Checks if the image can be expanded based on the alignment."""
+    if alignment in ("Left", "Right") and source_width >= target_width:
+        return False
+    if alignment in ("Top", "Bottom") and source_height >= target_height:
+        return False
+    return True
 
 @spaces.GPU
-def infer(image, model_selection, width, height, overlap_width, num_inference_steps, resize_size, prompt_input=None):
+def infer(image, width, height, overlap_width, num_inference_steps, resize_size, prompt_input=None, alignment="Middle"):
     source = image
     target_size = (width, height)
-    target_ratio = (width, height)
     overlap = overlap_width
 
     # New resizing logic based on resize_size
@@ -76,25 +78,63 @@ def infer(image, model_selection, width, height, overlap_width, num_inference_st
             new_height = int(source.height * scale_factor)
             source = source.resize((new_width, new_height), Image.LANCZOS)
 
-    margin_x = (target_size[0] - source.width) // 2
-    margin_y = (target_size[1] - source.height) // 2
+    if not can_expand(source.width, source.height, target_size[0], target_size[1], alignment):
+        alignment = "Middle"
+
+    # Calculate margins based on alignment
+    if alignment == "Middle":
+        margin_x = (target_size[0] - source.width) // 2
+        margin_y = (target_size[1] - source.height) // 2
+    elif alignment == "Left":
+        margin_x = 0
+        margin_y = (target_size[1] - source.height) // 2
+    elif alignment == "Right":
+        margin_x = target_size[0] - source.width
+        margin_y = (target_size[1] - source.height) // 2
+    elif alignment == "Top":
+        margin_x = (target_size[0] - source.width) // 2
+        margin_y = 0
+    elif alignment == "Bottom":
+        margin_x = (target_size[0] - source.width) // 2
+        margin_y = target_size[1] - source.height
 
     background = Image.new('RGB', target_size, (255, 255, 255))
     background.paste(source, (margin_x, margin_y))
 
     mask = Image.new('L', target_size, 255)
     mask_draw = ImageDraw.Draw(mask)
-    mask_draw.rectangle([
-        (margin_x + overlap, margin_y + overlap),
-        (margin_x + source.width - overlap, margin_y + source.height - overlap)
-    ], fill=0)
+
+    # Adjust mask generation based on alignment
+    if alignment == "Middle":
+        mask_draw.rectangle([
+            (margin_x + overlap, margin_y + overlap),
+            (margin_x + source.width - overlap, margin_y + source.height - overlap)
+        ], fill=0)
+    elif alignment == "Left":
+        mask_draw.rectangle([
+            (margin_x, margin_y),
+            (margin_x + source.width - overlap, margin_y + source.height)
+        ], fill=0)
+    elif alignment == "Right":
+        mask_draw.rectangle([
+            (margin_x + overlap, margin_y),
+            (margin_x + source.width, margin_y + source.height)
+        ], fill=0)
+    elif alignment == "Top":
+        mask_draw.rectangle([
+            (margin_x, margin_y),
+            (margin_x + source.width, margin_y + source.height - overlap)
+        ], fill=0)
+    elif alignment == "Bottom":
+        mask_draw.rectangle([
+            (margin_x, margin_y + overlap),
+            (margin_x + source.width, margin_y + source.height)
+        ], fill=0)
 
     cnet_image = background.copy()
     cnet_image.paste(0, (0, 0), mask)
 
-    final_prompt = "high quality"
-    if prompt_input.strip() != "":
-        final_prompt += ", " + prompt_input
+    final_prompt = f"{prompt_input} , high quality, 4k"
 
     (
         prompt_embeds,
@@ -118,7 +158,12 @@ def infer(image, model_selection, width, height, overlap_width, num_inference_st
 
     yield background, cnet_image
 
-def preload_presets(target_ratio):
+def clear_result():
+    """Clears the result ImageSlider."""
+    return gr.update(value=None)
+
+def preload_presets(target_ratio, ui_width, ui_height):
+    """Updates the width and height sliders based on the selected aspect ratio."""
     if target_ratio == "9:16":
         changed_width = 720
         changed_height = 1280
@@ -128,11 +173,15 @@ def preload_presets(target_ratio):
         changed_height = 720
         return changed_width, changed_height, gr.update(open=False)
     elif target_ratio == "Custom":
-        return 720, 1280, gr.update(open=True)
-
-def clear_result():
-    return gr.update(value=None)
+        return ui_width, ui_height, gr.update(open=True)
 
+def select_the_right_preset(user_width, user_height):
+    if user_width == 720 and user_height == 1280:
+        return "9:16"
+    elif user_width == 1280 and user_height == 720:
+        return "16:9"
+    else:
+        return "Custom"
 
 css = """
 .gradio-container {
@@ -140,7 +189,6 @@ css = """
 }
 """
 
-
 title = """<h1 align="center">Diffusers Image Outpaint</h1>
 <div align="center">Drop an image you would like to extend, pick your expected ratio and hit Generate.</div>
 <div style="display: flex; justify-content: center; align-items: center; text-align: center;">
@@ -160,69 +208,69 @@ with gr.Blocks(css=css) as demo:
             with gr.Column():
                 input_image = gr.Image(
                     type="pil",
-                    label="Input Image",
-                    sources=["upload"],
-                    height = 300
+                    label="Input Image"
                 )
-                
-                prompt_input = gr.Textbox(label="Prompt (Optional)")
-                
+
+                with gr.Row():
+                    with gr.Column(scale=2):
+                        prompt_input = gr.Textbox(label="Prompt (Optional)")
+                    with gr.Column(scale=1):
+                        run_button = gr.Button("Generate")
+
                 with gr.Row():
                     target_ratio = gr.Radio(
-                        label = "Expected Ratio",
-                        choices = ["9:16", "16:9", "Custom"],
-                        value = "9:16",
-                        scale = 2
+                        label="Expected Ratio",
+                        choices=["9:16", "16:9", "Custom"],
+                        value="9:16",
+                        scale=2
                     )
                     
-                    run_button = gr.Button("Generate", scale=1)
+                    alignment_dropdown = gr.Dropdown(
+                        choices=["Middle", "Left", "Right", "Top", "Bottom"],
+                        value="Middle",
+                        label="Alignment"
+                    )
 
                 with gr.Accordion(label="Advanced settings", open=False) as settings_panel:
-                    with gr.Column(): 
+                    with gr.Column():
                         with gr.Row():
                             width_slider = gr.Slider(
                                 label="Width",
                                 minimum=720,
-                                maximum=1440,
+                                maximum=1536,
                                 step=8,
                                 value=720,
                             )
                             height_slider = gr.Slider(
                                 label="Height",
                                 minimum=720,
-                                maximum=1440,
+                                maximum=1536,
                                 step=8,
                                 value=1280,
                             )
                         with gr.Row():
-                            model_selection = gr.Dropdown(
-                                choices=list(MODELS.keys()),
-                                value="RealVisXL V5.0 Lightning",
-                                label="Model",
-                            )
                             num_inference_steps = gr.Slider(label="Steps", minimum=4, maximum=12, step=1, value=8)
-
-                        overlap_width = gr.Slider(
-                            label="Mask overlap width",
-                            minimum=1,
-                            maximum=50,
-                            value=42,
-                            step=1
-                        )
-                        
+                            overlap_width = gr.Slider(
+                                label="Mask overlap width",
+                                minimum=1,
+                                maximum=50,
+                                value=42,
+                                step=1
+                            )
                         resize_size = gr.Number(
                             label="Resize Size (-1 for default behavior)",
                             value=-1,
                             precision=0
                         )
-
+                            
                 gr.Examples(
                     examples=[
-                        ["./examples/example_1.webp", "RealVisXL V5.0 Lightning", 1280, 720],  
-                        ["./examples/example_2.jpg", "RealVisXL V5.0 Lightning", 720, 1280],  
-                        ["./examples/example_3.jpg", "RealVisXL V5.0 Lightning", 1024, 1024],  
+                        ["./examples/example_1.webp", 1280, 720, "Middle"],
+                        ["./examples/example_2.jpg", 1440, 810, "Left"],
+                        ["./examples/example_3.jpg", 1024, 1024, "Top"],
+                        ["./examples/example_3.jpg", 1024, 1024, "Bottom"],
                     ],
-                    inputs=[input_image, model_selection, width_slider, height_slider],
+                    inputs=[input_image, width_slider, height_slider, alignment_dropdown],
                 )
 
             with gr.Column():
@@ -230,21 +278,52 @@ with gr.Blocks(css=css) as demo:
                     interactive=False,
                     label="Generated Image",
                 )
+                use_as_input_button = gr.Button("Use as Input Image", visible=False)
 
+    def use_output_as_input(output_image):
+        """Sets the generated output as the new input image."""
+        return gr.update(value=output_image[1])
+
+    use_as_input_button.click(
+        fn=use_output_as_input,
+        inputs=[result],
+        outputs=[input_image]
+    )
+    
     target_ratio.change(
-        fn = preload_presets,
-        inputs = [target_ratio],
-        outputs = [width_slider, height_slider, settings_panel],
-        queue = False
+        fn=preload_presets,
+        inputs=[target_ratio, width_slider, height_slider],
+        outputs=[width_slider, height_slider, settings_panel],
+        queue=False
+    )
+
+    width_slider.change(
+        fn=select_the_right_preset,
+        inputs=[width_slider, height_slider],
+        outputs=[target_ratio],
+        queue=False
+    )
+
+    height_slider.change(
+        fn=select_the_right_preset,
+        inputs=[width_slider, height_slider],
+        outputs=[target_ratio],
+        queue=False
     )
+    
     run_button.click(
         fn=clear_result,
         inputs=None,
         outputs=result,
     ).then(
         fn=infer,
-        inputs=[input_image, model_selection, width_slider, height_slider, overlap_width, num_inference_steps, resize_size, prompt_input],
+        inputs=[input_image, width_slider, height_slider, overlap_width, num_inference_steps,
+                resize_size, prompt_input, alignment_dropdown],
         outputs=result,
+    ).then(
+        fn=lambda: gr.update(visible=True),
+        inputs=None,
+        outputs=use_as_input_button,
     )
 
     prompt_input.submit(
@@ -253,8 +332,13 @@ with gr.Blocks(css=css) as demo:
         outputs=result,
     ).then(
         fn=infer,
-        inputs=[input_image, model_selection, width_slider, height_slider, overlap_width, num_inference_steps, resize_size, prompt_input],
+        inputs=[input_image, width_slider, height_slider, overlap_width, num_inference_steps,
+                resize_size, prompt_input, alignment_dropdown],
         outputs=result,
+    ).then(
+        fn=lambda: gr.update(visible=True),
+        inputs=None,
+        outputs=use_as_input_button,
     )
 
-demo.queue(max_size=12).launch(share=False, show_error=True, show_api=False)
+demo.queue(max_size=12).launch(share=False)