Update app.py

app.py

@@ -12,13 +12,12 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 pipe = pipe.to(device)
 
 # Define the inference function
-def ghibli_transform(input_image, prompt="ghibli style", strength=0.75, guidance_scale=7.5):
+def ghibli_transform(input_image, prompt="ghibli style", strength=0.75, guidance_scale=7.5, num_steps=50):
     if input_image is None:
         raise gr.Error("No image uploaded! Please upload an image before clicking Transform.")
     
     # Process the input image (keep it as PIL Image)
     try:
-        # Ensure the image is in RGB and resized
         init_image = input_image.convert("RGB").resize((768, 768))
     except Exception as e:
         raise gr.Error(f"Failed to process image: {str(e)}")
@@ -27,10 +26,10 @@ def ghibli_transform(input_image, prompt="ghibli style", strength=0.75, guidance
     try:
         output = pipe(
             prompt=prompt,
-            image=init_image,  # Pass PIL Image directly
+            image=init_image,
             strength=strength,
             guidance_scale=guidance_scale,
-            num_inference_steps=50
+            num_inference_steps=num_steps  # Use the UI-provided value
         ).images[0]
     except Exception as e:
         raise gr.Error(f"Pipeline error: {str(e)}")
@@ -48,6 +47,7 @@ with gr.Blocks(title="Ghibli Diffusion Image Transformer") as demo:
             prompt = gr.Textbox(label="Prompt", value="ghibli style")
             strength = gr.Slider(0, 1, value=0.75, step=0.05, label="Strength (How much to transform)")
             guidance = gr.Slider(1, 20, value=7.5, step=0.5, label="Guidance Scale")
+            num_steps = gr.Slider(10, 100, value=50, step=5, label="Inference Steps (Higher = Better Quality, Slower)")
             submit_btn = gr.Button("Transform")
         with gr.Column():
             output_img = gr.Image(label="Ghibli-Style Output")
@@ -55,7 +55,7 @@ with gr.Blocks(title="Ghibli Diffusion Image Transformer") as demo:
     # Connect the button to the function
     submit_btn.click(
         fn=ghibli_transform,
-        inputs=[input_img, prompt, strength, guidance],
+        inputs=[input_img, prompt, strength, guidance, num_steps],
         outputs=output_img
     )