LCM_Inpaint_Outpaint_Comfy
ComfyUI custom nodes for inpainting/outpainting using the new latent consistency model (LCM)
Nodes (42)
Push ComfyUI output into the canvas tool
A manual float dial for the pack's settings
Better detail from the fast model, one node
The node that tells you an image's width and height
Hand a canvas image back to ComfyUI
Image Resize (ComfyUI Essentials)
Randomly reorder four images (and read this before trusting the seed)
A two-way image swap, with a naming twist
This pack's main inpaint/outpaint workhorse
Fast img2img with loopback and a T2I adapter
Img2img steered by a ControlNet
Image-prompted variations at 4 steps
The everything-node for inpaint
Reference + ControlNet + IP-Adapter in one inpaint node
Steer LCM with a reference, prompt optional
The pack's SD-Turbo img2img sampler
Image-prompting on the fast LCM pipeline
IP-Adapter plus ControlNet in one LCM loader
The one node that actually loads the LCM model here
An LCM pipeline with ControlNet bolted in
The whole LCM + ControlNet inpaint rig in one node
The no-frills LCM img2img loader
Style transfer and promptless edits, reference-style
Restyle a masked region by example, at LCM speed
The pack's sideways turn into SD-Turbo
The generate node that does everything in the Lora line
The same workhorse, but its size inputs take wires
Controlnet + reference + IP-Adapter, all on one generate node
Turns any diffusers SD 1.5 model into a fast LCM inpaint rig
The loader with an LCM on/off switch
Grow your canvas one side at a time
Drop a T2I-Adapter into the LCM pipeline
SDXL speed the pack's own way
An image loader that takes a plain path
Pull your canvas state into the graph
Save the result and tell the canvas app about it
The plain save node the pack's workflows use
Save the piece, tell the jigsaw app it's ready
The puzzle saver's near-identical sequel
The pack's SDXL generation node, quirks included
Three saved presets for the ipadapter node's four dials
The Stitch Node Is the Glue Nobody Searches For — Here's What It Actually Does
Important
Errors may occur on Windows due to paths being formatted for linux without accounting for Windows. Let me know if you face any error in the issues page.
LCM_Inpaint-Outpaint_Comfy
ComfyUI custom nodes for inpainting/outpainting using the new latent consistency model (LCM)
Mix Images
<img src='https://github.com/taabata/LCM_Inpaint_Outpaint_Comfy/blob/main/LCM/instructions.png'>Inpaint
<img src='https://github.com/taabata/LCM_Inpaint-Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-10-24%2022-38-53.png'>Outpaint
<img src='https://github.com/taabata/LCM_Inpaint-Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-10-24%2022-42-53.png'>Prompt Weighting
Note: Requires CPU inference (select CPU in LCMLoader Node). (facing error that i dont know how to fix when using GPU)
Add '+' for more effect and '-' for less effect. Adding more '+' or '-' increases the effects.
<img src='https://github.com/taabata/LCM_Inpaint-Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-10-26%2020-18-11.png'>FreeU
<img src='https://github.com/taabata/LCM_Inpaint-Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-10-26%2020-39-33.png'>ReferenceOnly
<img src='https://github.com/taabata/LCM_Inpaint-Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-10-29%2000-09-32.png'>Style Transfer
<img src='https://github.com/taabata/LCM_Inpaint-Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-10-29%2020-08-52.png'>Image Variations
<img src='https://github.com/taabata/LCM_Inpaint-Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-10-29%2002-33-23.png'>Promptless Outpainting
<img src='https://github.com/taabata/LCM_Inpaint-Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-10-29%2003-32-29.png'>Image Blending
<img src='https://github.com/taabata/LCM_Inpaint-Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-10-29%2022-46-43.png'>ControlNet/T2I Adapter
<img src='https://github.com/taabata/LCM_Inpaint_Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-11-02%2023-41-10.png'>T2IAdapter thanks to Michael Poutre https://github.com/M1kep
Place model folders inside 'ComfyUI/models/controlnet'
IP Adapter
<img src='https://github.com/taabata/LCM_Inpaint_Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-11-06%2020-37-20.png'>Place model folders inside 'ComfyUI/models/controlnet'
Canvas Inpaint/Outpaint/img2img
<img src='https://github.com/taabata/LCM_Inpaint_Outpaint_Comfy/blob/main/LCM/Screenshot%20from%202023-11-15%2002-51-12.png'>In your Terminal/cmd at the directory where your ComfyUI folder is:
cd ComfyUI/custom_nodes/LCM_Inpaint_Outpaint_Comfy/CanvasTool
python setup.py
How to Use:
Clone into custom_nodes folder inside your ComfyUI directory
git clone https://github.com/taabata/LCM_Inpaint-Outpaint_Comfy
Install requirements after changing directory to LCM_Inpaint-Outpaint_Comfy folder
cd LCM_Inpaint-Outpaint_Comfy
pip install -r requirements.txt
Download the model in diffusers format from https://huggingface.co/SimianLuo/LCM_Dreamshaper_v7/tree/main and place it inside model/diffusers folder in your ComfyUI directory. (The name of the model folder should be "LCM_Dreamshaper_v7")
Load the workflow by choosing the .json file for inpainting or outpainting.
Credits
nagolinc's img2img script
Michael Poutre https://github.com/M1kep T2IAdapters