Multi image support

app.py

@@ -6,16 +6,16 @@ import torch
 import devicetorch
 import gradio as gr
 import numpy as np
-# import spaces
 from PIL import Image
-
+from dfloat11 import DFloat11Model
+#from kontext_pipeline import FluxKontextPipeline
 from diffusers import FluxKontextPipeline
 from diffusers.utils import load_image
-from dfloat11 import DFloat11Model
 
+# Load Kontext model
 MAX_SEED = np.iinfo(np.int32).max
 
-pipe = FluxKontextPipeline.from_pretrained("fuliucansheng/FLUX.1-Kontext-dev-diffusers", torch_dtype=torch.bfloat16)
+pipe = FluxKontextPipeline.from_pretrained("fuliucansheng/FLUX.1-Kontext-dev-diffusers", torch_dtype=torch.bfloat16).to("cuda")
 DFloat11Model.from_pretrained(
     "DFloat11/FLUX.1-Kontext-dev-DF11",
     device="cpu",
@@ -23,69 +23,116 @@ DFloat11Model.from_pretrained(
 )
 pipe.enable_model_cpu_offload()
 
-# @spaces.GPU
-def infer(input_image, prompt, seed=42, randomize_seed=False, guidance_scale=2.5, steps=28, progress=gr.Progress(track_tqdm=True)):
+def concatenate_images(images, direction="horizontal"):
     """
-    Perform image editing using the FLUX.1 Kontext pipeline.
-    
-    This function takes an input image and a text prompt to generate a modified version
-    of the image based on the provided instructions. It uses the FLUX.1 Kontext model
-    for contextual image editing tasks.
+    Concatenate multiple PIL images either horizontally or vertically.
     
     Args:
-        input_image (PIL.Image.Image): The input image to be edited. Will be converted
-            to RGB format if not already in that format.
-        prompt (str): Text description of the desired edit to apply to the image.
-            Examples: "Remove glasses", "Add a hat", "Change background to beach".
-        seed (int, optional): Random seed for reproducible generation. Defaults to 42.
-            Must be between 0 and MAX_SEED (2^31 - 1).
-        randomize_seed (bool, optional): If True, generates a random seed instead of
-            using the provided seed value. Defaults to False.
-        guidance_scale (float, optional): Controls how closely the model follows the
-            prompt. Higher values mean stronger adherence to the prompt but may reduce
-            image quality. Range: 1.0-10.0. Defaults to 2.5.
-        steps (int, optional): Controls how many steps to run the diffusion model for.
-            Range: 1-30. Defaults to 28.
-        progress (gr.Progress, optional): Gradio progress tracker for monitoring
-            generation progress. Defaults to gr.Progress(track_tqdm=True).
+        images: List of PIL Images
+        direction: "horizontal" or "vertical"
     
     Returns:
-        tuple: A 3-tuple containing:
-            - PIL.Image.Image: The generated/edited image
-            - int: The seed value used for generation (useful when randomize_seed=True)
-            - gr.update: Gradio update object to make the reuse button visible
-    
-    Example:
-        >>> edited_image, used_seed, button_update = infer(
-        ...     input_image=my_image,
-        ...     prompt="Add sunglasses",
-        ...     seed=123,
-        ...     randomize_seed=False,
-        ...     guidance_scale=2.5
-        ... )
+        PIL Image: Concatenated image
     """
+    if not images:
+        return None
+    
+    # Filter out None images
+    valid_images = [img for img in images if img is not None]
+    
+    if not valid_images:
+        return None
+    
+    if len(valid_images) == 1:
+        return valid_images[0].convert("RGB")
+    
+    # Convert all images to RGB
+    valid_images = [img.convert("RGB") for img in valid_images]
+    
+    if direction == "horizontal":
+        # Calculate total width and max height
+        total_width = sum(img.width for img in valid_images)
+        max_height = max(img.height for img in valid_images)
+        
+        # Create new image
+        concatenated = Image.new('RGB', (total_width, max_height), (255, 255, 255))
+        
+        # Paste images
+        x_offset = 0
+        for img in valid_images:
+            # Center image vertically if heights differ
+            y_offset = (max_height - img.height) // 2
+            concatenated.paste(img, (x_offset, y_offset))
+            x_offset += img.width
+            
+    else:  # vertical
+        # Calculate max width and total height
+        max_width = max(img.width for img in valid_images)
+        total_height = sum(img.height for img in valid_images)
+        
+        # Create new image
+        concatenated = Image.new('RGB', (max_width, total_height), (255, 255, 255))
+        
+        # Paste images
+        y_offset = 0
+        for img in valid_images:
+            # Center image horizontally if widths differ
+            x_offset = (max_width - img.width) // 2
+            concatenated.paste(img, (x_offset, y_offset))
+            y_offset += img.height
+    
+    return concatenated
+
+def infer(input_images, prompt, seed=42, randomize_seed=False, guidance_scale=4.0, steps=25, progress=gr.Progress(track_tqdm=True)):
+    
     if randomize_seed:
         seed = random.randint(0, MAX_SEED)
     
-    if input_image:
-        input_image = input_image.convert("RGB")
-        image = pipe(
-            image=input_image, 
-            prompt=prompt,
-            guidance_scale=guidance_scale,
-            width = input_image.size[0],
-            height = input_image.size[1],
-            num_inference_steps=steps,
-            generator=torch.Generator().manual_seed(seed),
-        ).images[0]
-    else:
-        image = pipe(
-            prompt=prompt,
-            guidance_scale=guidance_scale,
-            num_inference_steps=steps,
-            generator=torch.Generator().manual_seed(seed),
-        ).images[0]
+    # Handle input_images - it could be a single image or a list of images
+    if input_images is None:
+        raise gr.Error("Please upload at least one image.")
+    
+    # If it's a single image (not a list), convert to list
+    if not isinstance(input_images, list):
+        input_images = [input_images]
+    
+    # Filter out None images
+    valid_images = [img[0] for img in input_images if img is not None]
+    
+    if not valid_images:
+        raise gr.Error("Please upload at least one valid image.")
+    
+    # Concatenate images horizontally
+    concatenated_image = concatenate_images(valid_images, "horizontal")
+    
+    if concatenated_image is None:
+        raise gr.Error("Failed to process the input images.")
+    
+    # original_width, original_height = concatenated_image.size
+    
+    # if original_width >= original_height:
+    #     new_width = 1024
+    #     new_height = int(original_height * (new_width / original_width))
+    #     new_height = round(new_height / 64) * 64
+    # else:
+    #     new_height = 1024
+    #     new_width = int(original_width * (new_height / original_height))
+    #     new_width = round(new_width / 64) * 64
+    
+    #concatenated_image_resized = concatenated_image.resize((new_width, new_height), Image.LANCZOS)
 
+    final_prompt = f"From the provided reference images, create a unified, cohesive image such that {prompt}. Maintain the identity and characteristics of each subject while adjusting their proportions, scale, and positioning to create a harmonious, naturally balanced composition. Blend and integrate all elements seamlessly with consistent lighting, perspective, and style.the final result should look like a single naturally captured scene where all subjects are properly sized and positioned relative to each other, not assembled from multiple sources."
+    
+    image = pipe(
+        image=concatenated_image, 
+        prompt=final_prompt,
+        guidance_scale=guidance_scale,
+        width=concatenated_image.size[0],
+        height=concatenated_image.size[1],
+        num_inference_steps=steps,
+        generator=torch.Generator().manual_seed(seed),
+    ).images[0]
+    
     gradio_temp_dir = os.environ.get('GRADIO_TEMP_DIR', tempfile.gettempdir())
     temp_file_path = os.path.join(gradio_temp_dir, "image.png")
     image.save(temp_file_path, format="PNG")
@@ -94,14 +141,7 @@ def infer(input_image, prompt, seed=42, randomize_seed=False, guidance_scale=2.5
     gc.collect()
     devicetorch.empty_cache(torch)
 
-    return image, temp_file_path, seed, gr.Button(visible=True)
-
-# @spaces.GPU
-def infer_example(input_image, prompt):
-    image, temp_file_path, seed, _ = infer(input_image, prompt)
-    gc.collect()
-    devicetorch.empty_cache(torch)
-    return image,temp_file_path, seed
+    return image, seed, gr.update(visible=True)
 
 css="""
 #col-container {
@@ -114,7 +154,6 @@ css="""
 #row {
     min-height: 40vh; !Important
 }
-
 #row-height {
     height: 65px !important
 }
@@ -123,17 +162,26 @@ css="""
 with gr.Blocks(css=css) as demo:
     
     with gr.Column(elem_id="col-container"):
-        gr.Markdown(f"""# FLUX.1 Kontext [dev]
-Image editing and manipulation model guidance-distilled from FLUX.1 Kontext [pro], [[blog]](https://bfl.ai/announcements/flux-1-kontext-dev) [[model]](https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev)
+        gr.Markdown(f"""# FLUX.1 Kontext [dev] - Multi-Image
+        Flux Kontext with multiple image input support - compose a new image with elements from multiple images using Kontext [dev]
         """)
         with gr.Row(equal_height=True):
             with gr.Column():
-                input_image = gr.Image(label="Upload the image for editing", type="pil", elem_classes="input-image", elem_id="row")
-                    
+                input_images = gr.Gallery(
+                    label="Upload image(s) for editing", 
+                    show_label=True,
+                    elem_id="gallery_input",
+                    columns=3,
+                    rows=2,
+                    object_fit="contain",
+                    height="auto",
+                    file_types=['image'],
+                    type='pil'
+                )
+                
             with gr.Column():
                 result = gr.Image(label="Result", show_label=False, interactive=False, elem_classes="input-image", elem_id="row")
-                reuse_button = gr.Button("Reuse this image", visible=False)
-        
+
         with gr.Row(equal_height=True):
             with gr.Column():
                 prompt = gr.Text(
@@ -145,14 +193,14 @@ Image editing and manipulation model guidance-distilled from FLUX.1 Kontext [pro
                     container=True,
                     scale=1
                 )
-
+                
             with gr.Column():
-                download_image = gr.File(label="Download Image", elem_id="row-height", scale=0)
+                download_image = gr.File(label="Download Image", elem_id="row-height", interactive=False, scale=0)    
                 run_button = gr.Button("Run", scale=1)
 
         with gr.Row():
             with gr.Accordion("Advanced Settings", open=False):
-                
+        
                 seed = gr.Slider(
                     label="Seed",
                     minimum=0,
@@ -168,39 +216,30 @@ Image editing and manipulation model guidance-distilled from FLUX.1 Kontext [pro
                     minimum=1,
                     maximum=10,
                     step=0.1,
-                    value=2.5,
+                    value=4.0,
                 )       
                 
                 steps = gr.Slider(
                     label="Steps",
                     minimum=1,
                     maximum=40,
-                    value=28,
+                    value=25,
                     step=1
-                )
-            
-        examples = gr.Examples(
-            examples=[
-                ["flowers.png", "turn the flowers into sunflowers"],
-                ["monster.png", "make this monster ride a skateboard on the beach"],
-                ["cat.png", "make this cat happy"]
-            ],
-            inputs=[input_image, prompt],
-            outputs=[result, download_image, seed],
-            fn=infer_example,
-            cache_examples=False
-        )
-            
+                )       
+                    
+                reuse_button = gr.Button("Reuse this image", visible=False)
+        
     gr.on(
         triggers=[run_button.click, prompt.submit],
         fn = infer,
-        inputs = [input_image, prompt, seed, randomize_seed, guidance_scale, steps],
-        outputs = [result, download_image, seed, reuse_button]
+        inputs = [input_images, prompt, seed, randomize_seed, guidance_scale, steps],
+        outputs = [result, seed, reuse_button]
     )
+    
     reuse_button.click(
-        fn = lambda image: image,
+        fn = lambda image: [image] if image is not None else [],  # Convert single image to list for gallery
         inputs = [result],
-        outputs = [input_image]
+        outputs = [input_images]
     )
 
 demo.launch(mcp_server=True)