Hair style book (SD1.5)
5.0
0 reviewsDescription
the latest version update on 11/03:
Removed All Ollama Node Designs: Due to the limited number of usable hairstyles and the inability to accurately identify facial features to suggest appropriate hairstyles, the Ollama node design has been removed. The system now uses a preset hairstyle selection method instead.
Hairstyle Prompt Library with Claude: A hairstyle prompt library has been created using Claude, featuring 120 hairstyle configurations. Six hairstyles are randomly selected through Mixlab’s extraction commands, ensuring low repetition rates and eliminating the need for users to install Ollama or an LLM for prompt generation.
Simplified Translation Nodes: Translation now uses simple translation nodes to reduce computational resources.
In the previous version, due to local LLM limitations, prompts were generated randomly, often resulting in empty lines or disorganized outputs. This made it challenging to generate multiple images consistently, and hair masking was also unstable. In this update, users can draw their own mask.
Since AI struggles to handle bangs effectively, areas with bangs can also be excluded manually if needed. For manual masking, after selecting "Switch2," right-click on the node labeled "Switch2" to create the mask.
==
11/03改版
1、 移除了所有Ollama的节点设计,可用发型过少,也没有达成辨识图片人物脸型,建议发型的效果,实际几乎是随机抽选,故改为预置发型的方式。
2、 使用Claude建立发型提示库,包含发型配置方式,共计120种发型,透过Mixlab的抽取指令方式随机抽选6种,重复率低,使用者也毋需另行安装Ollama或LLM提词。
3、 翻译部分使用简单的翻译节点翻译,以降低运算资源。
前一版本因为本机LLM的缘故,提示词随机产生,分割后大多有空行或错乱,无法稳定产生多张图片,此外头发Mask也不稳定,此版本修改为可自行绘制Mask。
AI不擅长处理前额浏海,如果有前额浏海的部分,也可以手绘排除。
若有手绘Mask需求,选择Switch2后,右键单击在标有Switch2的节点上建立即可。
Discussion
(No comments yet)
Loading...
Reviews
No reviews yet
Versions (5)
- latest (10 months ago)
- v20240516-131133
- v20240516-115630
- v20240516-090524
- v20240516-065954
Node Details
Primitive Nodes (2)
Fast Groups Bypasser (rgthree) (1)
Note (1)
Custom Nodes (26)
- GoogleTranslateTextNode (1)
- CR Text Concatenate (1)
- CR Comic Panel Templates (1)
ComfyUI
- EmptyImage (1)
- PerturbedAttentionGuidance (1)
- PatchModelAddDownscale (1)
- PreviewImage (1)
- RandomNoise (1)
- LoadImage (1)
- SaveImage (1)
- ImageListToBatch+ (1)
- PreviewBridge (1)
- MaskToSEGS (1)
- ImpactSwitch (2)
- ToBasicPipe (1)
- DetailerForEachPipe (1)
- LayerStyle: DropShadow (1)
- LayerUtility: SimpleTextImage (1)
- LayerUtility: ImageBlendAdvance (1)
- LayerMask: PersonMaskUltra (1)
- RandomPrompt (1)
- Efficient Loader (1)
- ShowText|pysssss (1)
- String Literal (1)
- JWImageResizeByLongerSide (1)
Model Details
Checkpoints (0)
LoRAs (0)