comfy-capture-inference

A tool that captures the screen and infers via api to ComfyUI. Operation and display is done in Gradio.

Requirements

client (this repo)

$ pip install -r requirements.txt

server (ComfyUI)

# Setup https://github.com/comfyanonymous/ComfyUI
# Install below custom nodes.
# - https://github.com/Fannovel16/comfyui_controlnet_aux.git
# - https://github.com/Acly/comfyui-tooling-nodes
# Put
# - `Comfyui/models/loras/lcm-lora-sdv1-5.safetensors`
# - `Comfyui/models/controlnet/control_v11p_sd15_lineart.pth`
# - `Comfyui/models/controlnet/control_v11p_sd15_scribble.pth`

Usage

Describe the configuration in the config file. The model should be placed according to the ComfyUI configuration.

# config.json
{
    "init_workflow": "sd15_img2img.json",
    "ckpt_name": "sdhk_v40.safetensors",
    "seed": -1
}

# client
$ python main.py

# server (comfyUI)
$ python main.py

Custom Workflow

If a custom workflow is placed under a workflow, inference can be performed using that workflow. The following requirements must be met Please refer to the sample workflows.

• I/O nodes must be LoadImageBase64 and SendImageWebSocket.
• Each I/O node must be one.