uupdated nodes

.gitignore

@@ -42,5 +42,5 @@ comfy-venv
 comfy_venv_3.11
 models-2
 
-
 !comfy/ldm/models/autoencoder.py
+

custom_nodes/MemedeckComfyNodes/nodes_input.py

@@ -1,3 +1,4 @@
+import json
 from comfy_extras.nodes_custom_sampler import Noise_RandomNoise
 
 
@@ -16,7 +17,7 @@ class MD_VideoInputs:
                     "description": "The length of the video."
                 }),
                 "steps": ("INT", {
-                    "default": 25,
+                    "default": 15,
                     "description": "Number of steps to generate the video."
                 }),
                 "width": ("INT", {
@@ -58,10 +59,21 @@ class MD_VideoInputs:
             }
         }
 
-    RETURN_TYPES = ("STRING", "INT", "INT", "INT", "INT", "INT", "FLOAT", "STRING", "STRING", "STRING", "STRING", "NOISE",)
+    RETURN_TYPES = ("STRING", "INT", "INT", "INT", "INT", "INT", "FLOAT", "STRING", "STRING", "STRING", "STRING", "NOISE", "STRING")
-    RETURN_NAMES = ("image_url", "length", "steps", "width", "height", "crf", "terminal", "user_prompt", "pre_prompt", "post_prompt", "negative_prompt", "seed")
+    RETURN_NAMES = ("image_url", "length", "steps", "width", "height", "crf", "terminal", "user_prompt", "pre_prompt", "post_prompt", "negative_prompt", "seed", "input_metadata")
     FUNCTION = "load_inputs"
     CATEGORY = "MemeDeck"
 
-    def load_inputs(self, image_url, length=121, steps=25, width=768, height=768, crf=28, terminal=0.1, user_prompt="", pre_prompt="", post_prompt="", negative_prompt="", seed=None):
+    def load_inputs(self, image_url, length=121, steps=15, width=768, height=768, crf=28, terminal=0.1, user_prompt="", pre_prompt="", post_prompt="", negative_prompt="", seed=None):
-      return (image_url, length, steps, width, height, crf, terminal, user_prompt, pre_prompt, post_prompt, negative_prompt, Noise_RandomNoise(seed))
+        input_metadata = json.dumps({
+            "length": length,
+            "steps": steps,
+            "width": width,
+            "height": height,
+            "crf": crf,
+            "terminal": terminal,
+            "user_prompt": user_prompt,
+            "pre_prompt": pre_prompt,
+            "post_prompt": post_prompt,
+        })
+        return (image_url, length, steps, width, height, crf, terminal, user_prompt, pre_prompt, post_prompt, negative_prompt, Noise_RandomNoise(seed), input_metadata)

custom_nodes/MemedeckComfyNodes/nodes_output.py

@@ -47,11 +47,13 @@ class MD_SaveMP4:
                     "crf": ("FLOAT",),
                     "motion_prompt": ("STRING", ),
                     "negative_prompt": ("STRING", ),
-                    "img2vid_metadata": ("STRING", ),
+                },
-                    "sampler_metadata": ("STRING", ),
+                "optional": {
+                    "seed_value": ("NOISE", ),
+                    "input_metadata": ("STRING", ),
                 },
                 # "hidden": {"prompt": "PROMPT", "extra_pnginfo": "EXTRA_PNGINFO"},
-                }
+            }
 
     RETURN_TYPES = ()
     FUNCTION = "save_images"
@@ -60,7 +62,7 @@ class MD_SaveMP4:
 
     CATEGORY = "MemeDeck"
     
-    def save_images(self, images, fps, filename_prefix, crf=None, motion_prompt=None, negative_prompt=None, img2vid_metadata=None, sampler_metadata=None):
+    def save_images(self, images, fps, filename_prefix, crf=None, motion_prompt=None, negative_prompt=None, seed_value=None, input_metadata=None):
         start_time = time.time()
         filename_prefix += self.prefix_append
         full_output_folder, filename, counter, subfolder, filename_prefix = folder_paths.get_save_image_path(filename_prefix, self.output_dir, images[0].shape[1], images[0].shape[0])
@@ -82,14 +84,17 @@ class MD_SaveMP4:
 
         # metadata = pil_images[0].getexif()
         # num_frames = len(pil_images)
+        
+        logger.info(f"seed_value: {seed_value.seed}")
+        logger.info(f"input_metadata: {input_metadata}")
 
         json_metadata = {
             "crf": crf,
             "motion_prompt": motion_prompt,
             "negative_prompt": negative_prompt,
-            "img2vid_metadata": json.loads(img2vid_metadata),
+            "seed": seed_value.seed,
-            "sampler_metadata": json.loads(sampler_metadata),
+            "input_metadata": json.loads(input_metadata),
-        }            
+        }
 
         # Use ffmpeg to create MP4 with watermark
         output_file = f"{filename}_{counter:05}_.mp4"
@@ -237,105 +242,6 @@ class MD_SaveMP4:
 
         return watermark_img
     
-    # def save_images(self, images, fps, filename_prefix, lossless, quality, method, crf=None, motion_prompt=None, negative_prompt=None, img2vid_metadata=None, sampler_metadata=None):
-    #     start_time = time.time()
-    #     method = self.methods.get(method)
-    #     filename_prefix += self.prefix_append
-    #     full_output_folder, filename, counter, subfolder, filename_prefix = folder_paths.get_save_image_path(
-    #         filename_prefix, self.output_dir, images[0].shape[1], images[0].shape[0]
-    #     )
-    #     results = []
-
-    #     # Prepare PIL images in one loop
-    #     pil_images = [
-    #         Image.fromarray(np.clip((255. * image.cpu().numpy()), 0, 255).astype(np.uint8))
-    #         for image in images
-    #     ]
-
-    #     metadata = pil_images[0].getexif()
-    #     num_frames = len(pil_images)
-
-    #     # Pre-serialize JSON metadata
-    #     json_metadata = json.dumps({
-    #         "crf": crf,
-    #         "motion_prompt": motion_prompt,
-    #         "negative_prompt": negative_prompt,
-    #         "img2vid_metadata": json.loads(img2vid_metadata),
-    #         "sampler_metadata": json.loads(sampler_metadata),
-    #     })
-
-    #     # Save images directly
-    #     duration = int(1000.0 / fps)
-    #     for i in range(0, len(pil_images), num_frames):
-    #         file = f"{filename}_{counter:05}_.webp"
-    #         pil_images[i].save(
-    #             os.path.join(full_output_folder, file),
-    #             save_all=True,
-    #             duration=duration,
-    #             append_images=pil_images[i + 1:i + num_frames],
-    #             exif=metadata,
-    #             lossless=lossless,
-    #             quality=quality,
-    #             method=method
-    #         )
-    #         results.append({"filename": file, "subfolder": subfolder, "type": self.type})
-    #         counter += 1
-        
-    #     end_time = time.time()
-    #     logger.info(f"Save images took: {end_time - start_time} seconds")
-
-    #     return {
-    #         "ui": {
-    #             "images": results,
-    #             "animated": (num_frames != 1,),
-    #             "metadata": (json_metadata,),
-    #         },
-    #     }
-
-    # def save_images(self, images, fps, filename_prefix, lossless, quality, method, crf=None, motion_prompt=None, negative_prompt=None, img2vid_metadata=None, sampler_metadata=None):
-    #     method = self.methods.get(method)
-    #     filename_prefix += self.prefix_append
-    #     full_output_folder, filename, counter, subfolder, filename_prefix = folder_paths.get_save_image_path(filename_prefix, self.output_dir, images[0].shape[1], images[0].shape[0])
-    #     results = list()
-
-    #     pil_images = []
-    #     for image in images:
-    #         i = 255. * image.cpu().numpy()
-    #         img = Image.fromarray(np.clip(i, 0, 255).astype(np.uint8))
-    #         pil_images.append(img)
-
-    #     metadata = pil_images[0].getexif()
-    #     num_frames = len(pil_images)
-        
-    #     json_metadata = {
-    #       "crf": crf,
-    #       "motion_prompt": motion_prompt,
-    #       "negative_prompt": negative_prompt,
-    #       "img2vid_metadata": json.loads(img2vid_metadata),
-    #       "sampler_metadata": json.loads(sampler_metadata),
-    #     }
-
-    #     c = len(pil_images)
-    #     for i in range(0, c, num_frames):
-    #         file = f"{filename}_{counter:05}_.webp"
-    #         pil_images[i].save(os.path.join(full_output_folder, file), save_all=True, duration=int(1000.0/fps), append_images=pil_images[i + 1:i + num_frames], exif=metadata, lossless=lossless, quality=quality, method=method)
-    #         results.append({
-    #             "filename": file,
-    #             "subfolder": subfolder,
-    #             "type": self.type,
-    #         })
-    #         counter += 1
-
-    #     animated = num_frames != 1
-    #     # properly serialize metadata
-    #     return { 
-    #         "ui": { 
-    #             "images": results, 
-    #             "animated": (animated,), 
-    #             "metadata": (json.dumps(json_metadata),)
-    #         },
-    #     }
-    
 class MD_VAEDecode:
     @classmethod
     def INPUT_TYPES(s):

custom_nodes/MemedeckComfyNodes/nodes_preprocessing.py

@@ -1,3 +1,4 @@
+import json
 from pathlib import Path
 import sys
 import time
@@ -111,7 +112,11 @@ class MD_ImageToMotionPrompt:
                     },
                 ),
                 "max_tokens": ("INT", {"min": 1, "max": 2048, "default": 200}),
-            }
+            },
+            # "optional": {
+            #     "temperature": ("FLOAT", {"min": 0.0, "max": 1.0, "step": 0.01, "default": 0.2}),
+            #     "top_p": ("FLOAT", {"min": 0.0, "max": 1.0, "step": 0.01, "default": 0.9}),
+            # }
         }
 
     
@@ -121,14 +126,23 @@ class MD_ImageToMotionPrompt:
     CATEGORY = "MemeDeck"
 
     def generate_completion(
-        self, pre_prompt: str, post_prompt: str, Image: torch.Tensor, clip, prompt: str, negative_prompt: str, max_tokens: int
+        self, pre_prompt: str, post_prompt: str, Image: torch.Tensor, clip, prompt: str, negative_prompt: str,
+        # temperature: float, 
+        # top_p: float, 
+        max_tokens: int
     ) -> Tuple[str]:
         # start a timer
         start_time = time.time()
         b64image = image.pil2base64(image.tensor2pil(Image))
         # change this to a endpoint on localhost:5010/inference that takes a json with the image and the prompt
         
-        response = requests.post("http://127.0.0.1:5010/inference", json={"image_url": f"data:image/jpeg;base64,{b64image}", "prompt": prompt})
+        response = requests.post("http://127.0.0.1:5010/inference", json={
+            "image_url": f"data:image/jpeg;base64,{b64image}",
+            "prompt": prompt,
+            "temperature": 0.2,
+            "top_p": 0.7,
+            "max_gen_len": max_tokens,
+        })
         if response.status_code != 200:
             raise Exception(f"Failed to generate completion: {response.text}")
         end_time = time.time()
@@ -171,6 +185,19 @@ class MD_CompressAdjustNode:
                     "description": "The height of the video."
                 }),
             },
+            "optional": {
+                "weights": ("STRING", {
+                    "multiline": True,
+                    "default": json.dumps({
+                        "ideal_blockiness": 600,
+                        "ideal_edge_density": 12,
+                        "ideal_color_variation": 10000,
+                        "blockiness_weight": -0.006,
+                        "edge_density_weight": 0.32,
+                        "color_variation_weight": -0.00005
+                    }),
+                }),
+            }
         }
 
     RETURN_TYPES = ("IMAGE", "FLOAT", "INT", "INT")
@@ -252,11 +279,20 @@ class MD_CompressAdjustNode:
         target_crf = round(target_crf, 2)
         return target_crf
     
-    def tensor_to_video_and_back(self, image, desired_crf=28, width=832, height=832):        
+    def tensor_to_video_and_back(self, image, desired_crf=28, width=832, height=832, weights=None):        
         temp_dir = "temp_video"
         filename = f"frame_{time.time()}".split('.')[0]
         os.makedirs(temp_dir, exist_ok=True)
-                
+        
+        if weights:
+            weights = json.loads(weights)
+            self.ideal_blockiness = weights["ideal_blockiness"]
+            self.ideal_edge_density = weights["ideal_edge_density"]
+            self.ideal_color_variation = weights["ideal_color_variation"]
+            self.blockiness_weight = weights["blockiness_weight"]
+            self.edge_density_weight = weights["edge_density_weight"]
+            self.color_variation_weight = weights["color_variation_weight"]
+        
         # Convert single image to list if necessary
         if len(image.shape) == 3:
             image = [image]