ymc_node_joy

ymc_node_joy
★ 0

图像描述批量处理AIGCLoRA训练
在 ComfyUI 中用 Joy 为图像生成描述（caption）的自定义节点，支持单张与目录批量处理，便于 AIGC 管线与 LoRA 训练的数据准备
💡 为图像批量生成 Joy 风格的 caption 用于数据准备
🍴 3 Forks💻 Python🔄 2025-06-19
📦
网盘下载
复制链接后前往夸克网盘下载
https://pan.quark.cn/s/51137d50651f
📦 requirements.txt
#
Core
dependencies
#
torch==2.5.0+cu124
#
transformers==4.47.1
#
numpy==1.26.3
#
Pillow==10.1.0
#
torch
#
transformers
#
numpy
#
Pillow
#
ComfyUI
specific
#
#
Custom
node
utilities
#
yors_comfyui_node_setup==0.10.1
#
yors_pano_ansi_color==1.2.1
yors_comfyui_node_setup
yors_pano_ansi_color
#
Optional
dependencies
(for
advanced
features)
#
bitsandbytes==0.45.0
#
bitsandbytes
📄 README

ymc_node_joy

🤖 comfyui custom nodes to caption image with joy

Why

  • use joy to caption image for aigc.
  • use joy to caption image files in diretory to do sth. (eg. lora training)

    • Features

    Nodes

  • nodes show in console:

  • nodes show in right mouse menu:

    • Install

    # cd to comfyui/custom_nodes
git clone https://github.com/ymc-github/ymc_node_joy
  • deps will be installed automatically if deps in requirements.txt were not installed when comfyui up

    • Required Models

    Essential components:

  • google/siglip-so400m-patch14-384 (Vision model)
  • unsloth/Meta-Llama-3.1-8B-bnb-4bit or meta-llama/Meta-Llama-3.1-8B (LLM)
  • Joy_caption/image_adapter.pt (Custom adapter)

    • <comfyui_root>/
├── models/
│   ├── clip/                    # SigLIP Vision Model
│   │   └── siglip-so400m-patch14-384/
│   ├── llm/                     # Llama Language Model
│   │   ├── Meta-Llama-3.1-8B-bnb-4bit/
│   │   └── Meta-Llama-3.1-8B/
│   └── Joy_caption/             # Custom Components
│       └── image_adapter.pt     # Dimension Adapter

    1. SigLIP Vision Model (google/siglip-so400m-patch14-384)

    International: https://huggingface.co/google/siglip-so400m-patch14-384

    China Mirror: https://hf-mirror.com/google/siglip-so400m-patch14-384

    2. Llama Language Models

    4bit Quantized (unsloth/Meta-Llama-3.1-8B-bnb-4bit)

    International: https://huggingface.co/unsloth/Meta-Llama-3.1-8B-bnb-4bit

    China Mirror: https://hf-mirror.com/unsloth/Meta-Llama-3.1-8B-bnb-4bit

    Original (meta-llama/Meta-Llama-3.1-8B)

    International: https://huggingface.co/meta-llama/Meta-Llama-3.1-8B (Access approval required)

    China Mirror: https://hf-mirror.com/meta-llama/Meta-Llama-3.1-8B

    3. Image Adapter (Joy_caption/image_adapter.pt)

    International: https://huggingface.co/spaces/fancyfeast/joy-caption-pre-alpha/tree/main/wpkklhc6

    China Mirror: https://www.modelscope.cn/models/fireicewolf/joy-caption-pre-alpha/files

    Usage

  • you can find it in search box : double click + typing keywords (eg: joy, caption)
  • you can find it in right mouse menu : ymc/caption
  • ~~you can find it in right mouse menu : utils/ymc/caption (as alias)~~

    • Demo

  • workflow demo:

    • Based-on

  • pypi package yors_comfyui_node_setup – setup comfyui custom nodes easily
  • pypi package yors_pano_ansi_color – info msg in console with color in your comfyui custom nodes easily
  • ~~pypi package yors_pano_zero_field – set nodes input field to be HQ in your comfyui custom nodes easily~~

    • Published to Comfy registry

  • get more details in publish_to_comfy.yml

  • docs for publishing to comfy registey

  • installed with comfy-cli ? comfy node registry-install ymc_node_joy

    • Author

    name|email|desciption

    :–|:–|:–

    yemiancheng||Main developer and code maintainer|

    chenxinghua|<455758525@qq.com>|Code reference from StartHua/Comfyui_CXH_joy_caption|

    License

    MIT