release version

.gitignore

@@ -1,2 +1,3 @@
 __pycache__
 .idea
+.DS_Store

README.md

@@ -1,4 +1,122 @@
-# ComfyUI-FaceChain
-continue coding....
-Expected to be released on January 20
-预计1月20日左右发布，不会鸽
+# [ComfyUI-FaceChain](https://github.com/THtianhao/ComfyUI-FaceChain)
+
+This project is adapted from [facechain](https://github.com/modelscope/facechain) and involves the breakdown and improvement of the processes of [facechain](https://github.com/modelscope/facechain).
+
+![workflow_inpaiting_inference.png](workflows%2Fworkflow_inpaiting_inference.png)
+
+English | [简体中文](./README_zh-CN.md)
+
+## Contact
+
+If you have any questions or suggestions, you can reach us through the following channels:
+
+- Email: tototianhao@gmail.com
+- Telegram: https://t.me/+JoFE2vqHU4phZjg1
+- WeChat Group: <img src="./images/wechat.jpg" width="300">
+
+### Steps
+1. Install ComfyUI first.
+
+2. After a successful installation of ComfyUI, navigate to the `custom_nodes` directory at `ComfyUI/custom_nodes/`.
+
+```
+cd custom_nodes
+```
+
+3. Clone this project into the `custom_nodes` directory.
+
+```
+git clone https://github.com/THtianhao/ComfyUI-Portrait-Maker.git
+```
+
+4. Restart ComfyUI.
+
+## ComfyUI Workflow
+
+Facechain Workflow Location: [workflow_inpaiting_inference.png](workflows%2Fworkflow_inpaiting_inference.png)
+Drag the workflow directly into ComfyUI.
+
+## Node Introduction
+### FC StyleLoraLoad
+  > The workflow can load the checkpoints and style Lora used by facechain, download them first, and then merge them, providing relevant prompts.
+  > Workflow: [./workflow/workflow_inference.json](./workflows/workflow_inference.json)
+### FC FaceDetectCrop
+  > Detects faces and crops them.
+   
+ ![workflow_face_detect_crop.png](workflows%2Fworkflow_face_detect_crop.png) 
+  Parameter Description:
+
+  1. mode: Cropping mode, normal mode crops according to the face, square 512 width height will scale the face to 512.
+  2. face_index: Index of the face, if there are multiple faces, retrieve them based on the index.
+  3. crop_ratio: Only effective in normal mode, crops the face proportionally, 1.0 is 1x the face.
+
+### FC FaceFusion
+  > Fusion using model scope model.
+ 
+ ![workflow_face_fusion.png](workflows%2Fworkflow_face_fusion.png) 
+
+  
+### FC FaceSegment
+  > Segmentation using model scope model to obtain masks for the face and body.
+
+![workflow_face_segment.png](workflows%2Fworkflow_face_segment.png)  
+  
+Parameter Description:
+1. ksize: Expansion parameter for segmenting the edges of the face.
+2. ksize1: Expansion parameter for segmenting the edges of the face.
+3. include_neck: Whether the segmented image includes the neck.
+
+### FC FaceSegAndReplace
+  > Performs face fusion and replaces the original image, similar to facefusion but mainly used for multiple people.
+
+![workflow_face_seg_and_replace.png](workflows%2Fworkflow_face_seg_and_replace.png)  
+
+
+### FC RemoveCannyFace
+  > Removes the Canny-detected parts of the face.
+  
+![workflow_remove_canny_face.png](workflows%2Fworkflow_remove_canny_face.png)
+### FC ReplaceByMask
+  > Replaces the image based on the mask.
+  
+![workflow_replace_by_mask.png](workflows%2Fworkflow_replace_by_mask.png)
+
+* FC MaskOP
+  > Operations on the mask.
+
+Parameter Description:
+1. mode: Provides three operations, blur, erosion, dilation.
+2. kernel: The kernel used for the operation, the larger the kernel, the stronger the operation.
+  
+![workflow_mask_op.png](workflows%2Fworkflow_mask_op.png)
+
+* FC FCCropToOrigin
+  > Currently, it can only be used in conjunction with `FC FaceDetectCrop` in `square 512 width height` mode. Pastes the cropped image onto the target image based on the mask.
+  
+![workflow_face_detect_crop.png](workflows%2Fworkflow_face_detect_crop.png)
+
+Parameter Description:
+1. origin_image: Original image.
+2. origin_box: Bounding box of the original image.
+3. origin_mask: Mask cropped from the original image.
+4. paste_image: Pasting image, must be consistent with the origin_mask, hence the need for `FC FaceDetectCrop` in `square 512 width height` mode.
+  
+
+
+
+## Contribution
+
+If you find any issues or have suggestions for improvement, feel free to contribute. Follow these steps:
+
+1. Branch out a new feature branch: `git checkout -b feature/your-feature-name`
+2. Make your changes and commit: `git commit -m "Add new feature"`
+3. Push to your remote branch: `git push origin feature/your-feature-name`
+4. Create a Pull Request (PR).
+
+## License
+
+This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.
+
+
+
+Join us and contribute to the development of EasyPhoto ComfyUI Plugin!

README_zh-CN.md

@@ -0,0 +1,122 @@
+# [ComfyUI-FaceChain](https://github.com/THtianhao/ComfyUI-FaceChain)
+
+项目改编于[facechain](https://github.com/modelscope/facechain)，对于[facechain](https://github.com/modelscope/facechain)进行了流程上的拆解和改进。
+
+![workflow_inpaiting_inference.png](workflows%2Fworkflow_inpaiting_inference.png)
+
+English | [简体中文](./README_zh-CN.md)
+
+## 联系
+
+如果你有任何疑问或建议，可以通过以下方式联系我们：
+
+- 电子邮件：tototianhao@gmail.com
+- telegram: https://t.me/+JoFE2vqHU4phZjg1
+- 微信群： <img src="./images/wechat.jpg" width="300">
+
+### 步骤
+1. 首先安装ComfyUI
+
+2. ComfyUI运行成功后进入`custom_nodes` 目录 `ComfyUI/custom_nodes/`
+
+```
+cd custom_nodes
+```
+
+3. 克隆此项目到custom_nodes目录中
+
+```
+git clone https://github.com/THtianhao/ComfyUI-Portrait-Maker.git
+```
+
+4. 重新启动ComfyUI
+
+## ComfyUI 工作流
+
+Facechain工作位置: [workflow_inpaiting_inference.png](workflows%2Fworkflow_inpaiting_inference.png)
+将工作流直接拖进comfyui中
+
+## 节点介绍
+### FC StyleLoraLoad
+  > workflow可以加载facechain 用到的checkpoint和风格lora，首先下载然后进行融合，并且给出相关的prompt
+  > workflow ： [./workflow/workflow_inference.json](./workflows/workflow_inference.json)
+### FC FaceDetectCrop
+  > 识别人脸并且剪裁
+   
+ ![workflow_face_detect_crop.png](workflows%2Fworkflow_face_detect_crop.png) 
+  参数介绍：
+
+  1. mode:剪裁模式，normal模式为按照人脸剪裁,square 512 width height会将人脸缩放到512
+  2. face_index:人脸的索引，如果有多个人脸的话按照多个人脸的话按照index进行获取
+  3. crop_ratio:只在normal模式下生效，将人脸按照比例剪裁1.0为1倍人脸
+
+### FC FaceFusion
+  > 使用model scope模型进行融合
+ 
+ ![workflow_face_fusion.png](workflows%2Fworkflow_face_fusion.png) 
+
+  
+### FC FaceSegment
+  > 使用model scope模型进行分割并且获取脸部和身体的mask
+
+![workflow_face_segment.png](workflows%2Fworkflow_face_segment.png)  
+  
+参数介绍:
+1. ksize: 分割人脸边缘的扩展参数
+2. ksize1: 分割人脸边缘的扩展参数
+3. include_neck: 分割的图像是否包含脖子
+
+### FC FaceSegAndReplace
+  > 进行人脸融合并且分割人脸替换原图,差异和facefusion不大,主要用在多人
+
+![workflow_face_seg_and_replace.png](workflows%2Fworkflow_face_seg_and_replace.png)  
+
+
+### FC RemoveCannyFace
+  > 删除掉canny的人脸部分
+  
+![workflow_remove_canny_face.png](workflows%2Fworkflow_remove_canny_face.png)
+### FC ReplaceByMask
+  > 根据mask替换图像
+  
+![workflow_replace_by_mask.png](workflows%2Fworkflow_replace_by_mask.png)
+
+* FC MaskOP
+  > 对mask的操作
+
+参数介绍:
+1. mode: 提供了三种操作，模糊处理，腐蚀，膨胀
+2. kernel: 用于操作的核，越大操作力度越强
+  
+![workflow_mask_op.png](workflows%2Fworkflow_mask_op.png)
+
+* FC FCCropToOrigin
+  > 目前只能配合 `FC FaceDetectCrop` 的`square 512 width height`模式一起使用,将截取的图像根据mask粘贴到目标图像上
+  
+![workflow_face_detect_crop.png](workflows%2Fworkflow_face_detect_crop.png)
+
+参数介绍:
+1. origin_image: 原始图像
+2. origin_box:原始图像的bbox
+3. origin_mask:原始图像截取的mask
+4. paste_image:粘贴图像 必须和origin_mask保持一致,因此需要`FC FaceDetectCrop` 的`square 512 width height`模式
+  
+
+
+
+## 贡献
+
+如果你发现任何问题或有改进建议，欢迎贡献。请遵循以下步骤：
+
+1. 分支出一个新的特性分支：`git checkout -b feature/your-feature-name`
+2. 进行修改并提交：`git commit -m "Add new feature"`
+3. 推送到你的远程分支：`git push origin feature/your-feature-name`
+4. 创建一个 Pull 请求（PR）。
+
+## 许可证
+
+该项目采用 MIT 许可证。查看 [LICENSE](LICENSE) 文件以获取详细信息。
+
+
+
+欢迎加入我们，为 EasyPhoto ConfyUI Plugin 的发展做出贡献！

facechain/common/model_processor.py

@@ -48,7 +48,7 @@ def facechain_detect_crop(source_image_pil, face_index=0, crop_ratio=1, mode='no
         debug('mask', mask.shape)
         corp_img_pil = source_image_pil.crop(bbox)
         return corp_img_pil, mask, bbox, points_array
-    elif mode == "square 512 width heigh":
+    elif mode == "square 512 width height":
         np_image = image_to_np(source_image_pil)
         face_ratio = 0.45
         crop_l = int(max(face_h, face_w) / face_ratio / 2)
@@ -69,6 +69,7 @@ def facechain_detect_crop(source_image_pil, face_index=0, crop_ratio=1, mode='no
         raise RuntimeError('模式错误')
 
 def segment(img, ksize=0, eyeh=0, ksize1=0, include_neck=False, warp_mask=None, return_human=True):
+    print('ksize', ksize, "ksize1", ksize1, 'warp_mask', warp_mask)
     seg_image = get_segmentation()(img)
     masks = seg_image['masks']
     scores = seg_image['scores']

facechain/nodes.py

@@ -32,7 +32,7 @@ class FaceDetectCrop:
                 "source_image": ("IMAGE",),
                 "face_index": ("INT", {"default": 0, "min": 0, "max": 10, "step": 1}),
                 "crop_ratio": ("FLOAT", {"default": 1.0, "min": 0, "max": 10, "step": 0.1}),
-                "mode": (["normal", "square 512 width heigh"],),
+                "mode": (["normal", "square 512 width height"],),
             }
         }
 
@@ -54,8 +54,8 @@ class FCFaceSegment:
 
             },
             "optional": {
-                "ksize": ("FLOAT", {"default": 0, "min": 0, "max": 10, "step": 0.1}),
-                "ksize1": ("FLOAT", {"default": 0, "min": 0, "max": 10, "step": 0.1}),
+                "ksize": ("FLOAT", {"default": 0, "min": 0, "max": 1, "step": 0.01}),
+                "ksize1": ("FLOAT", {"default": 0, "min": 0, "max": 1, "step": 0.01}),
                 "include_neck": ("BOOLEAN", {"default": False, "label_on": "enabled", "label_off": "disabled"}),
                 "warp_mask": ("MASK",),
             },
@@ -66,9 +66,9 @@ class FCFaceSegment:
     FUNCTION = "fc_segment"
     CATEGORY = "facechain/model"
 
-    def fc_segment(self, source_image, ksize=0, ksize1=0, include_neck=False, warp_mask=None, ):
+    def fc_segment(self, source_image, ksize, ksize1, include_neck=False, warp_mask=None, ):
         pil_source_image = tensor_to_img(source_image)
-        seg_image, mask, human_mask = segment(pil_source_image, ksize, ksize1, include_neck, warp_mask, True)
+        seg_image, mask, human_mask = segment(pil_source_image, ksize=ksize, ksize1=ksize1, include_neck=include_neck, warp_mask=warp_mask, return_human=True)
         return image_to_tensor(seg_image), mask_np2_to_mask_tensor(mask), mask_np2_to_mask_tensor(human_mask)
 
 class FCFaceFusionAndSegReplace:
@@ -177,7 +177,7 @@ class FCReplaceByMask:
         np_source_image = tensor_to_np(source_image)
         np_replace_image = tensor_to_np(replace_image)
         np_mask = mask_tensor_to_mask_np3(mask)
-        result_np = np_source_image * np_mask + np_replace_image(1 - np_mask)
+        result_np = np_source_image * np_mask + np_replace_image * (1 - np_mask)
         return (image_np_to_image_tensor(result_np),)
 
 class FCCropAndPaste:
@@ -213,8 +213,8 @@ class FCMaskOP:
         return {
             "required": {
                 "mask": ("MASK",),
-                "ksize": ("INT", {"default": 8, "min": 0, "max": 100, "step": 1}),
-                "method": (["expand_dims", "concatenate", "burl", "erode"],),
+                "method": (["burl", "erode", "dilate"],),
+                "kernel": ("INT", {"default": 16, "min": 0, "max": 100, "step": 1}),
             }
         }
 
@@ -222,19 +222,19 @@ class FCMaskOP:
     FUNCTION = "mask_op"
     CATEGORY = "facechain/mask"
 
-    def mask_op(self, mask, ksize, method):
+    def mask_op(self, mask, method, kernel):
         mask_np = mask_tensor_to_mask_np3(mask)
         result = None
-        if method == "concatenate":
-            result = np.concatenate([mask_np, mask_np, mask_np], axis=2)
-        elif method == "expand_dims":
-            result = np.expand_dims(mask_np, axis=2)
-        elif method == "burl":
-            result = cv2.GaussianBlur(mask_np, (int(ksize * 1.8) * 2 + 1, int(ksize * 1.8) * 2 + 1), 0)
+        kernel_real = np.ones((kernel, kernel), np.uint8)
+        if method == "burl":
+            result = cv2.GaussianBlur(mask_np, (int(kernel * 1.8) * 2 + 1, int(kernel * 1.8) * 2 + 1), 0)
             result = np.expand_dims(result, axis=2)
         elif method == 'erode':
-            kernel = np.ones((ksize * 2, ksize * 2))
-            result = cv2.erode(mask_np, kernel, iterations=1)
+            result = cv2.erode(mask_np, kernel_real, iterations=1)
+            result = np.expand_dims(result, axis=2)
+        elif method == 'dilate':
+            result = cv2.dilate(mask_np, kernel_real, iterations=1)
+            result = np.expand_dims(result, axis=2)
         return (mask_np3_to_mask_tensor(result),)
 
 class FCCropToOrigin:

facechain/style_loader_node.py

@@ -54,7 +54,7 @@ class FCStyleLoraLoad:
             }
         }
 
-    RETURN_TYPES = ("MODEL", "CLIP", "VAE", "CONDITION",)
+    RETURN_TYPES = ("MODEL", "CLIP", "VAE", "STRING",)
     RETURN_NAMES = ("MODEL", "CLIP", "VAE", "style_prompt",)
     FUNCTION = "style_lora_load"
     CATEGORY = "facechain/lora"