added tab and func changes

app.py

@@ -177,8 +177,17 @@ def extract_frames(video_path):
     cap.release()
     return frames
 
+#process_video_frames
 @spaces.GPU(duration=120)
-def process_video_frames(video, cloth_image, cloth_type, num_inference_steps, guidance_scale, seed, show_type):
+def process_video_frames(
+        video,
+        cloth_image,
+        cloth_type,
+        num_inference_steps,
+        guidance_scale,
+        seed,
+        show_type
+):
     """
     Process each frame of the video through the flux pipeline
     
@@ -196,34 +205,16 @@ def process_video_frames(video, cloth_image, cloth_type, num_inference_steps, gu
     processed_frames = []
     
     for person_image in frames:
-        # Create a generator if seed is set
-        generator = None
-        if seed != -1:
-            generator = torch.Generator(device='cuda').manual_seed(seed)
-        
-        # Resize and process person image
-        person_image = resize_and_crop(person_image, (args.width, args.height))
-        
-        # Load and resize cloth image (do this outside the loop if it's constant)
-        cloth_image = Image.open(cloth_image).convert("RGB")
-        cloth_image = resize_and_padding(cloth_image, (args.width, args.height))
-        
-        # Generate mask (you might want to optimize this if mask is similar across frames)
-        mask = automasker(person_image, cloth_type)['mask']
-        mask = mask_processor.blur(mask, blur_factor=9)
-        
-        # Run inference
-        result_image = pipeline_flux(
-            image=person_image,
-            condition_image=cloth_image,
-            mask_image=mask,
-            width=args.width,
-            height=args.height,
-            num_inference_steps=num_inference_steps,
-            guidance_scale=guidance_scale,
-            generator=generator
-        ).images[0]
-        
+        new_result_image, result_image = submit_function_flux(
+            person_image,
+            cloth_image,
+        cloth_type,
+        num_inference_steps,
+        guidance_scale,
+        seed,
+        show_type
+        )
+        yield result_image
         processed_frames.append(result_image)
     
     yield processed_frames
@@ -399,65 +390,71 @@ def app_gradio():
             with gr.Row():
                 with gr.Column(scale=1, min_width=350):
                     with gr.Row():
-                        image_path_p2p = gr.Image(
+                        image_path_vidflux = gr.Image(
                             type="filepath",
                             interactive=True,
                             visible=False,
                         )
-                        person_image_p2p = gr.Video(
-                           label="Person Image",
+                        person_image_vidflux = gr.Video(
+                            
                         )
 
                     with gr.Row():
                         with gr.Column(scale=1, min_width=230):
-                            cloth_image_p2p = gr.Image(
+                            cloth_image_vidflux = gr.Image(
                                 interactive=True, label="Condition Image", type="filepath"
                             )
+                        with gr.Column(scale=1, min_width=120):
+                            gr.Markdown(
+                                '<span style="color: #808080; font-size: small;">Two ways to provide Mask:<br>1. Upload the person image and use the `🖌️` above to draw the Mask (higher priority)<br>2. Select the `Try-On Cloth Type` to generate automatically </span>'
+                            )
+                            cloth_type = gr.Radio(
+                                label="Try-On Cloth Type",
+                                choices=["upper", "lower", "overall"],
+                                value="upper",
+                            )
 
-                    submit_p2p = gr.Button("Submit")
+                    submit_flux = gr.Button("Submit")
                     gr.Markdown(
                         '<center><span style="color: #FF0000">!!! Click only Once, Wait for Delay !!!</span></center>'
                     )
 
-                    gr.Markdown(
-                        '<span style="color: #808080; font-size: small;">Advanced options can adjust details:<br>1. `Inference Step` may enhance details;<br>2. `CFG` is highly correlated with saturation;<br>3. `Random seed` may improve pseudo-shadow.</span>'
-                    )
                     with gr.Accordion("Advanced Options", open=False):
-                        num_inference_steps_p2p = gr.Slider(
+                        num_inference_steps_vidflux = gr.Slider(
                             label="Inference Step", minimum=10, maximum=100, step=5, value=50
                         )
                         # Guidence Scale
-                        guidance_scale_p2p = gr.Slider(
-                            label="CFG Strenth", minimum=0.0, maximum=7.5, step=0.5, value=2.5
+                        guidance_scale_vidflux = gr.Slider(
+                            label="CFG Strenth", minimum=0.0, maximum=50, step=0.5, value=30
                         )
                         # Random Seed
-                        seed_p2p = gr.Slider(
+                        seed_vidflux = gr.Slider(
                             label="Seed", minimum=-1, maximum=10000, step=1, value=42
                         )
-                        # show_type = gr.Radio(
-                        #     label="Show Type",
-                        #     choices=["result only", "input & result", "input & mask & result"],
-                        #     value="input & mask & result",
-                        # )
+                        show_type = gr.Radio(
+                            label="Show Type",
+                            choices=["result only", "input & result", "input & mask & result"],
+                            value="input & mask & result",
+                        )
 
                 with gr.Column(scale=2, min_width=500):
-                    #result_image_p2p = gr.Image(interactive=False, label="Result")
-                    gall_output = gr.Gallery(label="Processed Frames")
-
+                    result_image_vidflux = gr.Image(interactive=False, label="Result")
+                    with gr.Row():
+                        # Photo Examples
+                        root_path = "resource/demo/example"
+                        with gr.Column():
+                            gal_output = gr.Gallery(label="Processed Frames")
+                        
 
-                image_path_p2p.change(
-                    person_example_fn, inputs=image_path_p2p, outputs=person_image_p2p
+                image_path_vidflux.change(
+                    person_example_fn, inputs=image_path_vidflux, outputs=person_image_vidflux
                 )
 
-                submit_p2p.click(
+                submit_flux.click(
                     process_video_frames,
-                    [
-                        person_image_p2p,
-                        cloth_image_p2p,
-                        num_inference_steps_p2p,
-                        guidance_scale_p2p,
-                        seed_p2p],
-                    gall_output,
+                    [person_image_vidflux, cloth_image_vidflux, cloth_type, num_inference_steps_vidflux, guidance_scale_vidflux,
+                     seed_vidflux, show_type],
+                    result_image_vidflux,gal_output
                 )