我有看到Comfy作者自己整了一个简单的workflow,就是那种以前的SD模型也通用的workflow,这种是不可以的,会生成非常垃圾甚至非常可怕的人像(四肢乱飞那种),一定要用SD官方在huggingface上传的那几个sample workflow(就是其中包含CLIP loader、好几个conditioning set timestep range还有那个model sampling sd3的node的复杂一...
https://openart.ai/workflows/oliviosarikas/mesh-graphormer-hand-fixing-controlnet/NkhzwEW80FzCcvzzXEsH 此工作流程使用HandRefiner和ControlNet来检测手部区域和深度,然后重绘,修复效果更好。 要正常运行这个工作流程,你需要在这里下载3个模型:hr16/ControlNet-HandRefiner-pruned at main 把control_sd15_inpaint_...
Playground:PlaygroundAI 发布的开源模型,官方声称其优于 SDXL和 Midjourney5.2,1024px 的基础尺寸。官方发布有可用于 SDUI 工具的 safetensors 模型,本文的封面图就是用 Playground_v2.5 加 ComfyUI 的远古提示词生成的,直出效果确实比 SDXL 好,它的使用方法与 SDXL 一样,有兴趣的同学可以尝试下,抱脸上搜索 ...
SDXL can also be used only with the base model. This can be useful if you have limited resources or experimenting in general. Experiments In the experiments we are applying what we've learned in the category to try to explore more advanced workflows. ...
A repository of well documented easy to follow workflows for ComfyUI - ComfyUI_Workflows/image_conditioning/experiments/unclip_SDXL_refiner.json at main · cubiq/ComfyUI_Workflows
App For Everyone from Creators to Entrepreneurs or use open-source version → Load Any AI Model You can use all the current open-source models available on the market with our product, such as SD 1.5, SDXL. Even some closed-source models such as Ideogram and Flux Pro....
📂将所有工作流保存并同步到本地文件夹中(默认情况下,在“设置”中的“自定义”下)/ComfyUI/my_workflows 通过按💾按钮或⌨Shift+S 快捷键(在“设置”中自定义)保存工作流程。 每次保存时,它都会在“版本历史记录”中创建一个新记录。您可以轻松恢复到任何版本。
Significant performance optimization for SDXL model inference High customizability, allowing users granular control Portable workflows that can be shared easily Developer-friendly Due to these advantages, ComfyUI is increasingly being used by artistic creators. In this post, we will introduce ho...
模型选择“ip-adapter instant id sdxl [eb2d3ec0l” 点击红圈中的“爆炸”按钮,会自动下载预处理文件。 再在第二个ControlNet单元1中选择上传一张面部图片,它会确定面部的位置,设置如下。 预处理器选择“instant id face keypoints” 模型选择“control instant id sdxl [c5c25a50]” ...
模型选择“ip-adapter instant id sdxl [eb2d3ec0l” 点击红圈中的“爆炸”按钮,会自动下载预处理文件。 再在第二个ControlNet单元1中选择上传一张面部图片,它会确定面部的位置,设置如下。 预处理器选择“instant id face keypoints” 模型选择“control instant id sdxl [c5c25a50]” ...