Update app.py

app.py

@@ -108,7 +108,7 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 pipe = QwenImageEditPlusPipeline.from_pretrained(
     "Qwen/Qwen-Image-Edit-2509",
     transformer=QwenImageTransformer2DModel.from_pretrained(
-        "linoyts/Qwen-Image-Edit-Rapid-AIO", # [transformer weights extracted from: Phr00t/Qwen-Image-Edit-Rapid-AIO]
+        "linoyts/Qwen-Image-Edit-Rapid-AIO",
         subfolder='transformer',
         torch_dtype=dtype,
         device_map='cuda'
@@ -140,13 +140,6 @@ pipe.load_lora_weights("wiikoo/Qwen-lora-nsfw",
 pipe.load_lora_weights("wiikoo/Qwen-lora-nsfw",
                        weight_name="loras/Qwen4Play_v2.safetensors",
                        adapter_name="upscale-image")
-# pipe.load_lora_weights("lovis93/next-scene-qwen-image-lora-2509",
-#                        weight_name="next-scene_lora-v2-3000.safetensors",
-#                        adapter_name="next-scene")
-# pipe.load_lora_weights("wiikoo/Qwen-lora-nsfw",
-#                        weight_name="loras/Qwen_Nsfw_Body_V14-10K.safetensors",
-#                        adapter_name="upscale-image")
-
 
 pipe.transformer.set_attn_processor(QwenDoubleStreamAttnProcessorFA3())
 MAX_SEED = np.iinfo(np.int32).max
@@ -226,22 +219,22 @@ def infer(
     ).images
 
     try:
-        # Подготовка команды для ffmpeg
+        # Prepare ffmpeg command
         cmd = [
             'ffmpeg',
-            '-i', 'pipe:0',           # Чтение входных данных из stdin
-            '-vf', 'gblur=sigma=25',  # Фильтр размытия
-            '-f', 'image2pipe',       # Формат вывода - изображение в pipe
-            '-c:v', 'png',            # Кодек для вывода (можно использовать 'bmp' или 'mjpeg')
-            'pipe:1'                  # Вывод результата в stdout
+            '-i', 'pipe:0',
+            '-vf', 'gblur=sigma=25',
+            '-f', 'image2pipe',
+            '-c:v', 'png',
+            'pipe:1'
         ]
         
-        # Конвертируем PIL Image в байты
+        # Convert PIL Image to bytes
         img_byte_arr = BytesIO()
-        images[0].save(img_byte_arr, format='PNG')  # Сохраняем в формате PNG
+        images[0].save(img_byte_arr, format='PNG')
         img_data = img_byte_arr.getvalue()
     
-        # Запускаем ffmpeg и передаем данные через pipe
+        # Run ffmpeg with pipe
         result = subprocess.run(
             cmd, 
             input=img_data, 
@@ -249,14 +242,11 @@ def infer(
             check=True
         )
         
-        # Читаем результат из stdout
+        # Read result from stdout
         blurred_image_data = BytesIO(result.stdout)
         blurred_image = Image.open(blurred_image_data)
         
-        # Теперь blurred_image - это PIL Image объект
-        # Далее вы можете сохранить его или использовать как нужно
-        
-        # Если всё же нужен временный файл для результата:
+        # Save blurred image to temp file
         with tempfile.NamedTemporaryFile(suffix=".jpg", delete=False) as blur_tmpfile:
             blurred_image.save(blur_tmpfile.name, 'JPEG')
             blurred_img_path = blur_tmpfile.name
@@ -264,16 +254,7 @@ def infer(
         return images[0], blurred_img_path, seed
     
     except Exception as e:
-        raise Exception(f"Ошибка блюра: {e}")
-
-@spaces.GPU(duration=60)
-def infer_example(input_image, prompt, lora_adapter):
-    input_pil = input_image.convert("RGB")
-    guidance_scale = 1.0
-    steps = 5
-    result, seed = infer(input_pil, prompt, lora_adapter, 0, True, guidance_scale, steps)
-    return result, seed
-
+        raise Exception(f"Blur error: {e}")
 
 css="""
 #col-container {
@@ -315,31 +296,6 @@ with gr.Blocks(css=css, theme=steel_blue_theme) as demo:
                     randomize_seed = gr.Checkbox(label="Randomize Seed", value=True)
                     guidance_scale = gr.Slider(label="Guidance Scale", minimum=1.0, maximum=10.0, step=0.1, value=1.0)
                     steps = gr.Slider(label="Inference Steps", minimum=1, maximum=50, step=1, value=4)
-        
-        gr.Examples(
-            examples=[
-                ["examples/1.jpg", "Transform into anime.", "Photo-to-Anime"],
-                ["examples/5.jpg", "Remove shadows and relight the image using soft lighting.", "Light-Restoration"],
-                ["examples/4.jpg", "Use a subtle golden-hour filter with smooth light diffusion.", "Relight"],
-                ["examples/2.jpeg", "Rotate the camera 45 degrees to the left.", "Multiple-Angles"],
-                ["examples/7.jpg", "Light source from the Right Rear", "Multi-Angle-Lighting"],
-                ["examples/10.jpeg", "Upscale the image.", "Upscale-Image"],
-                ["examples/7.jpg", "Light source from the Below", "Multi-Angle-Lighting"],
-                ["examples/2.jpeg", "Switch the camera to a top-down right corner view.", "Multiple-Angles"],
-                ["examples/9.jpg", "The camera moves slightly forward as sunlight breaks through the clouds, casting a soft glow around the character's silhouette in the mist. Realistic cinematic style, atmospheric depth.", "Next-Scene"],
-                ["examples/8.jpg", "Make the subjects skin details more prominent and natural.", "Edit-Skin"],
-                ["examples/6.jpg", "Switch the camera to a bottom-up view.", "Multiple-Angles"],
-                ["examples/6.jpg", "Rotate the camera 180 degrees upside down.", "Multiple-Angles"],
-                ["examples/4.jpg", "Rotate the camera 45 degrees to the right.", "Multiple-Angles"],
-                ["examples/4.jpg", "Switch the camera to a top-down view.", "Multiple-Angles"],
-                ["examples/4.jpg", "Switch the camera to a wide-angle lens.", "Multiple-Angles"],
-            ],
-            inputs=[input_image, prompt, lora_adapter],
-            outputs=[output_image1, output_image2, seed],
-            fn=infer_example,
-            cache_examples=False,
-            label="Examples"
-        )
 
     run_button.click(
         fn=infer,