@ignix-social/creator 0.0.0-stage → 0.1.0

package/dist/box-kit/comfyui/recast/wan_utils.py

@@ -0,0 +1,227 @@
+# Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved.
+# Modified by Ignix (2026-10-05) for Ignix Creator: file renamed, unchanged. Apache-2.0 (see LICENSE-WAN.txt).
+import os
+import cv2
+import math
+import random
+import numpy as np
+
+def get_mask_boxes(mask):
+    """
+
+    Args:
+        mask: [h, w]
+    Returns:
+
+    """
+    y_coords, x_coords = np.nonzero(mask)
+    x_min = x_coords.min()
+    x_max = x_coords.max()
+    y_min = y_coords.min()
+    y_max = y_coords.max()
+    bbox = np.array([x_min, y_min, x_max, y_max]).astype(np.int32)
+    return bbox
+
+
+def get_aug_mask(body_mask, w_len=10, h_len=20):
+    body_bbox = get_mask_boxes(body_mask)
+
+    bbox_wh = body_bbox[2:4] - body_bbox[0:2]
+    w_slice = np.int32(bbox_wh[0] / w_len)
+    h_slice = np.int32(bbox_wh[1] / h_len)
+
+    for each_w in range(body_bbox[0], body_bbox[2], w_slice):
+        w_start = min(each_w, body_bbox[2])
+        w_end = min((each_w + w_slice), body_bbox[2])
+        # print(w_start, w_end)
+        for each_h in range(body_bbox[1], body_bbox[3], h_slice):
+            h_start = min(each_h, body_bbox[3])
+            h_end = min((each_h + h_slice), body_bbox[3])
+            if body_mask[h_start:h_end, w_start:w_end].sum() > 0:
+                body_mask[h_start:h_end, w_start:w_end] = 1
+
+    return body_mask
+    
+def get_mask_body_img(img_copy, hand_mask, k=7, iterations=1):
+    kernel = np.ones((k, k), np.uint8)
+    dilation = cv2.dilate(hand_mask, kernel, iterations=iterations)
+    mask_hand_img = img_copy * (1 - dilation[:, :, None])
+
+    return mask_hand_img, dilation
+
+
+def get_face_bboxes(kp2ds, scale, image_shape, ratio_aug):
+    h, w = image_shape
+    kp2ds_face = kp2ds.copy()[23:91, :2]
+
+    min_x, min_y = np.min(kp2ds_face, axis=0)
+    max_x, max_y = np.max(kp2ds_face, axis=0)
+
+
+    initial_width = max_x - min_x
+    initial_height = max_y - min_y
+
+    initial_area = initial_width * initial_height
+
+    expanded_area = initial_area * scale
+
+    new_width = np.sqrt(expanded_area * (initial_width / initial_height))
+    new_height = np.sqrt(expanded_area * (initial_height / initial_width))
+
+    delta_width = (new_width - initial_width) / 2
+    delta_height = (new_height - initial_height) / 4
+
+    if ratio_aug:
+        if random.random() > 0.5:
+            delta_width += random.uniform(0, initial_width // 10)
+        else:
+            delta_height += random.uniform(0, initial_height // 10)
+
+    expanded_min_x = max(min_x - delta_width, 0)
+    expanded_max_x = min(max_x + delta_width, w)
+    expanded_min_y = max(min_y - 3 * delta_height, 0)
+    expanded_max_y = min(max_y + delta_height, h)
+
+    return [int(expanded_min_x), int(expanded_max_x), int(expanded_min_y), int(expanded_max_y)]
+
+
+def calculate_new_size(orig_w, orig_h, target_area, divisor=64):
+
+    target_ratio = orig_w / orig_h
+
+    def check_valid(w, h):
+
+        if w <= 0 or h <= 0:
+            return False
+        return (w * h <= target_area and  
+                w % divisor == 0 and  
+                h % divisor == 0)  
+
+    def get_ratio_diff(w, h):
+
+        return abs(w / h - target_ratio)
+
+    def round_to_64(value, round_up=False, divisor=64):
+
+        if round_up:
+            return divisor * ((value + (divisor - 1)) // divisor)
+        return divisor * (value // divisor)
+
+    possible_sizes = []
+
+    max_area_h = int(np.sqrt(target_area / target_ratio))
+    max_area_w = int(max_area_h * target_ratio)
+
+    max_h = round_to_64(max_area_h, round_up=True, divisor=divisor)
+    max_w = round_to_64(max_area_w, round_up=True, divisor=divisor)
+
+    for h in range(divisor, max_h + divisor, divisor):
+        ideal_w = h * target_ratio
+
+        w_down = round_to_64(ideal_w)
+        w_up = round_to_64(ideal_w, round_up=True)
+
+        for w in [w_down, w_up]:
+            if check_valid(w, h, divisor):
+                possible_sizes.append((w, h, get_ratio_diff(w, h)))
+
+    if not possible_sizes:
+        raise ValueError("Can not find suitable size")
+
+    possible_sizes.sort(key=lambda x: (-x[0] * x[1], x[2]))
+
+    best_w, best_h, _ = possible_sizes[0]
+    return int(best_w), int(best_h)
+
+
+def resize_by_area(image, target_area, keep_aspect_ratio=True, divisor=64, padding_color=(0, 0, 0)):
+    h, w = image.shape[:2]
+    try:
+        new_w, new_h = calculate_new_size(w, h, target_area, divisor)
+    except:
+        aspect_ratio = w / h
+
+        if keep_aspect_ratio:
+            new_h = math.sqrt(target_area / aspect_ratio)
+            new_w = target_area / new_h
+        else:
+            new_w = new_h = math.sqrt(target_area)
+
+        new_w, new_h = int((new_w // divisor) * divisor), int((new_h // divisor) * divisor)
+
+    interpolation = cv2.INTER_AREA if (new_w * new_h < w * h) else cv2.INTER_LINEAR
+
+    resized_image = padding_resize(image, height=new_h, width=new_w, padding_color=padding_color,
+                                    interpolation=interpolation)
+    return resized_image
+
+
+def padding_resize(img_ori, height=512, width=512, padding_color=(0, 0, 0), interpolation=cv2.INTER_LINEAR):
+    ori_height = img_ori.shape[0]
+    ori_width = img_ori.shape[1]
+    channel = img_ori.shape[2]
+
+    img_pad = np.zeros((height, width, channel))
+    if channel == 1:
+        img_pad[:, :, 0] = padding_color[0]
+    else:
+        img_pad[:, :, 0] = padding_color[0]
+        img_pad[:, :, 1] = padding_color[1]
+        img_pad[:, :, 2] = padding_color[2]
+
+    if (ori_height / ori_width) > (height / width):
+        new_width = int(height / ori_height * ori_width)
+        img = cv2.resize(img_ori, (new_width, height), interpolation=interpolation)
+        padding = int((width - new_width) / 2)
+        if len(img.shape) == 2:
+            img = img[:, :, np.newaxis]  
+        img_pad[:, padding: padding + new_width, :] = img
+    else:
+        new_height = int(width / ori_width * ori_height)
+        img = cv2.resize(img_ori, (width, new_height), interpolation=interpolation)
+        padding = int((height - new_height) / 2)
+        if len(img.shape) == 2:
+            img = img[:, :, np.newaxis]  
+        img_pad[padding: padding + new_height, :, :] = img
+
+    img_pad = np.uint8(img_pad)
+
+    return img_pad
+
+
+def get_frame_indices(frame_num, video_fps, clip_length, train_fps):
+
+    start_frame = 0
+    times = np.arange(0, clip_length) / train_fps
+    frame_indices = start_frame + np.round(times * video_fps).astype(int)
+    frame_indices = np.clip(frame_indices, 0, frame_num - 1)
+
+    return frame_indices.tolist()
+
+
+def get_face_bboxes(kp2ds, scale, image_shape):
+    h, w = image_shape
+    kp2ds_face = kp2ds.copy()[1:] * (w, h)
+
+    min_x, min_y = np.min(kp2ds_face, axis=0)
+    max_x, max_y = np.max(kp2ds_face, axis=0)
+
+    initial_width = max_x - min_x
+    initial_height = max_y - min_y
+
+    initial_area = initial_width * initial_height
+
+    expanded_area = initial_area * scale
+
+    new_width = np.sqrt(expanded_area * (initial_width / initial_height))
+    new_height = np.sqrt(expanded_area * (initial_height / initial_width))
+
+    delta_width = (new_width - initial_width) / 2
+    delta_height = (new_height - initial_height) / 4
+
+    expanded_min_x = max(min_x - delta_width, 0)
+    expanded_max_x = min(max_x + delta_width, w)
+    expanded_min_y = max(min_y - 3 * delta_height, 0)
+    expanded_max_y = min(max_y + delta_height, h)
+
+    return [int(expanded_min_x), int(expanded_max_x), int(expanded_min_y), int(expanded_max_y)]

package/dist/box-kit/comfyui/workflows/infinitetalk.v1.json

@@ -0,0 +1,94 @@
+{
+  "1": {
+    "class_type": "UNETLoader",
+    "inputs": {
+      "unet_name": "Wan2_1-I2V-14B-480p_fp8_e4m3fn_scaled_KJ.safetensors",
+      "weight_dtype": "default"
+    }
+  },
+  "2": {
+    "class_type": "LoraLoaderModelOnly",
+    "inputs": {
+      "model": ["1", 0],
+      "lora_name": "lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors",
+      "strength_model": 1
+    }
+  },
+  "3": {
+    "class_type": "CLIPLoader",
+    "inputs": {
+      "clip_name": "umt5_xxl_fp8_e4m3fn_scaled.safetensors",
+      "type": "wan",
+      "device": "default"
+    }
+  },
+  "4": { "class_type": "VAELoader", "inputs": { "vae_name": "Wan2_1_VAE_bf16.safetensors" } },
+  "5": {
+    "class_type": "ModelPatchLoader",
+    "inputs": { "name": "wan2.1_infiniteTalk_single_fp16.safetensors" }
+  },
+  "6": {
+    "class_type": "AudioEncoderLoader",
+    "inputs": { "audio_encoder_name": "wav2vec2-chinese-base_fp16.safetensors" }
+  },
+  "7": { "class_type": "LoadImage", "inputs": { "image": "{{image}}" } },
+  "8": { "class_type": "LoadAudio", "inputs": { "audio": "{{audio}}" } },
+  "9": {
+    "class_type": "AudioEncoderEncode",
+    "inputs": { "audio_encoder": ["6", 0], "audio": ["8", 0] }
+  },
+  "10": { "class_type": "CLIPTextEncode", "inputs": { "clip": ["3", 0], "text": "{{prompt}}" } },
+  "11": { "class_type": "ConditioningZeroOut", "inputs": { "conditioning": ["10", 0] } },
+  "12": {
+    "class_type": "WanInfiniteTalkToVideo",
+    "inputs": {
+      "mode": "single_speaker",
+      "model": ["2", 0],
+      "model_patch": ["5", 0],
+      "positive": ["10", 0],
+      "negative": ["11", 0],
+      "vae": ["4", 0],
+      "width": "{{width}}",
+      "height": "{{height}}",
+      "length": "{{frames}}",
+      "start_image": ["7", 0],
+      "audio_encoder_output_1": ["9", 0],
+      "motion_frame_count": 9,
+      "audio_scale": 1
+    }
+  },
+  "13": { "class_type": "KSamplerSelect", "inputs": { "sampler_name": "euler" } },
+  "14": {
+    "class_type": "CFGGuider",
+    "inputs": { "model": ["12", 0], "positive": ["12", 1], "negative": ["12", 2], "cfg": 1 }
+  },
+  "15": {
+    "class_type": "BasicScheduler",
+    "inputs": { "model": ["12", 0], "scheduler": "normal", "steps": 6, "denoise": 1 }
+  },
+  "16": { "class_type": "RandomNoise", "inputs": { "noise_seed": "{{seed}}" } },
+  "17": {
+    "class_type": "SamplerCustomAdvanced",
+    "inputs": {
+      "noise": ["16", 0],
+      "guider": ["14", 0],
+      "sampler": ["13", 0],
+      "sigmas": ["15", 0],
+      "latent_image": ["12", 3]
+    }
+  },
+  "18": { "class_type": "VAEDecode", "inputs": { "samples": ["17", 0], "vae": ["4", 0] } },
+  "19": {
+    "class_type": "CreateVideo",
+    "inputs": { "images": ["18", 0], "fps": 25, "audio": ["8", 0] }
+  },
+  "20": {
+    "class_type": "SaveVideo",
+    "inputs": {
+      "video": ["19", 0],
+      "filename_prefix": "creator/out",
+      "format": "auto",
+      "codec": "auto"
+    }
+  }
+}

package/dist/box-kit/comfyui/workflows/wan22-i2v-14b.v1.json

@@ -0,0 +1,112 @@
+{
+  "1": {
+    "class_type": "UNETLoader",
+    "inputs": {
+      "unet_name": "wan2.2_i2v_high_noise_14B_fp8_scaled.safetensors",
+      "weight_dtype": "default"
+    }
+  },
+  "2": {
+    "class_type": "UNETLoader",
+    "inputs": {
+      "unet_name": "wan2.2_i2v_low_noise_14B_fp8_scaled.safetensors",
+      "weight_dtype": "default"
+    }
+  },
+  "3": {
+    "class_type": "LoraLoaderModelOnly",
+    "inputs": {
+      "model": ["1", 0],
+      "lora_name": "wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors",
+      "strength_model": 1
+    }
+  },
+  "4": {
+    "class_type": "LoraLoaderModelOnly",
+    "inputs": {
+      "model": ["2", 0],
+      "lora_name": "wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors",
+      "strength_model": 1
+    }
+  },
+  "5": { "class_type": "ModelSamplingSD3", "inputs": { "model": ["3", 0], "shift": 5 } },
+  "6": { "class_type": "ModelSamplingSD3", "inputs": { "model": ["4", 0], "shift": 5 } },
+  "7": {
+    "class_type": "CLIPLoader",
+    "inputs": {
+      "clip_name": "umt5_xxl_fp8_e4m3fn_scaled.safetensors",
+      "type": "wan",
+      "device": "default"
+    }
+  },
+  "8": { "class_type": "CLIPTextEncode", "inputs": { "clip": ["7", 0], "text": "{{prompt}}" } },
+  "9": {
+    "class_type": "CLIPTextEncode",
+    "inputs": {
+      "clip": ["7", 0],
+      "text": "static, frozen, blurry, low quality, jpeg artifacts, distorted, deformed hands, extra fingers, warped face, flicker, jitter, camera shake, new people entering, hands entering the frame, new objects appearing, text, subtitles, watermark, logo"
+    }
+  },
+  "10": { "class_type": "VAELoader", "inputs": { "vae_name": "Wan2_1_VAE_bf16.safetensors" } },
+  "11": { "class_type": "LoadImage", "inputs": { "image": "{{image}}" } },
+  "12": {
+    "class_type": "WanImageToVideo",
+    "inputs": {
+      "positive": ["8", 0],
+      "negative": ["9", 0],
+      "vae": ["10", 0],
+      "width": "{{width}}",
+      "height": "{{height}}",
+      "length": "{{frames}}",
+      "batch_size": 1,
+      "start_image": ["11", 0]
+    }
+  },
+  "13": {
+    "class_type": "KSamplerAdvanced",
+    "inputs": {
+      "model": ["5", 0],
+      "add_noise": "enable",
+      "noise_seed": "{{seed}}",
+      "steps": 4,
+      "cfg": 1,
+      "sampler_name": "euler",
+      "scheduler": "simple",
+      "positive": ["12", 0],
+      "negative": ["12", 1],
+      "latent_image": ["12", 2],
+      "start_at_step": 0,
+      "end_at_step": 2,
+      "return_with_leftover_noise": "enable"
+    }
+  },
+  "14": {
+    "class_type": "KSamplerAdvanced",
+    "inputs": {
+      "model": ["6", 0],
+      "add_noise": "disable",
+      "noise_seed": 0,
+      "steps": 4,
+      "cfg": 1,
+      "sampler_name": "euler",
+      "scheduler": "simple",
+      "positive": ["12", 0],
+      "negative": ["12", 1],
+      "latent_image": ["13", 0],
+      "start_at_step": 2,
+      "end_at_step": 10000,
+      "return_with_leftover_noise": "disable"
+    }
+  },
+  "15": { "class_type": "VAEDecode", "inputs": { "samples": ["14", 0], "vae": ["10", 0] } },
+  "16": { "class_type": "CreateVideo", "inputs": { "images": ["15", 0], "fps": 24 } },
+  "17": {
+    "class_type": "SaveVideo",
+    "inputs": {
+      "video": ["16", 0],
+      "filename_prefix": "creator/out",
+      "format": "auto",
+      "codec": "auto"
+    }
+  }
+}

package/dist/box-kit/comfyui/workflows/wan22-i2v.v1.json

@@ -0,0 +1,62 @@
+{
+  "1": {
+    "class_type": "UNETLoader",
+    "inputs": { "unet_name": "wan2.2_ti2v_5B_fp16.safetensors", "weight_dtype": "default" }
+  },
+  "2": {
+    "class_type": "CLIPLoader",
+    "inputs": {
+      "clip_name": "umt5_xxl_fp8_e4m3fn_scaled.safetensors",
+      "type": "wan",
+      "device": "default"
+    }
+  },
+  "3": { "class_type": "VAELoader", "inputs": { "vae_name": "wan2.2_vae.safetensors" } },
+  "4": { "class_type": "ModelSamplingSD3", "inputs": { "model": ["1", 0], "shift": 8 } },
+  "5": { "class_type": "CLIPTextEncode", "inputs": { "clip": ["2", 0], "text": "{{prompt}}" } },
+  "6": {
+    "class_type": "CLIPTextEncode",
+    "inputs": {
+      "clip": ["2", 0],
+      "text": "static, frozen, blurry, low quality, jpeg artifacts, distorted, deformed hands, extra fingers, warped face, flicker, jitter, camera shake, new people entering, hands entering the frame, new objects appearing, text, subtitles, watermark, logo"
+    }
+  },
+  "7": { "class_type": "LoadImage", "inputs": { "image": "{{image}}" } },
+  "8": {
+    "class_type": "Wan22ImageToVideoLatent",
+    "inputs": {
+      "vae": ["3", 0],
+      "width": "{{width}}",
+      "height": "{{height}}",
+      "length": "{{frames}}",
+      "batch_size": 1,
+      "start_image": ["7", 0]
+    }
+  },
+  "9": {
+    "class_type": "KSampler",
+    "inputs": {
+      "model": ["4", 0],
+      "positive": ["5", 0],
+      "negative": ["6", 0],
+      "latent_image": ["8", 0],
+      "seed": "{{seed}}",
+      "steps": 20,
+      "cfg": 5,
+      "sampler_name": "uni_pc",
+      "scheduler": "simple",
+      "denoise": 1
+    }
+  },
+  "10": { "class_type": "VAEDecode", "inputs": { "samples": ["9", 0], "vae": ["3", 0] } },
+  "11": { "class_type": "CreateVideo", "inputs": { "images": ["10", 0], "fps": 24 } },
+  "12": {
+    "class_type": "SaveVideo",
+    "inputs": {
+      "video": ["11", 0],
+      "filename_prefix": "creator/out",
+      "format": "auto",
+      "codec": "auto"
+    }
+  }
+}