The Impact Pack supports image enhancement through inpainting using Detector, Detailer, and Bridge nodes, offering various workflow configuration methods through Wildcards, Regional Sampler, Logics, PIPE, and more.
Please refer to the GitHub page for more detailed information.
https://github.com/ltdrdata/ComfyUI-Impact-Pack
Install guide:
Method 1 (strongly recommended):
Install through ComfyUI-Manager
Method 2:
Download
Uncompress into ComfyUI/custom_nodes
GOOD: ComfyUI/custom_nodes/ComfyUI-Impact-Pack
BAD CASE 1: with redundant subfolder
ComfyUI/custom_nodes/ComfyUI-Impact-Pack/ComfyUI-Impact-Pack
BAD CASE 2: without subfolder
ComfyUI/custom_nodes
Restart ComfyUI
Method 3:
goto ComfyUI/custom_nodes in cmd
Restart ComfyUI
Troubleshootings:
Occasionally, when a new parameter is created in an update, the values of nodes created in the previous version can be shifted to different fields. This can result in unintended results or errors if executed as is, so it is important to check the node values. In particular, when updating from version v1.4 or earlier to version v1.5, all parameter values under guide_size will be different, so be careful when using previously created workflows.
Tutorial videos:
https://www.youtube.com/channel/UCmML5IfnB_06Mx_MsWmnzSg
Tutorial resources and worksflows:
https://github.com/ltdrdata/ComfyUI-extension-tutorials
Updates:
v4.52.1
Detailer: AnimateDiff support, noise/unsampler hook, wildcards
Preview Bridge (latent)
Logics nodes (Switches, ControlBridge, QueueTrigger, ..)
Regional Sampler
And so on....
v3.20
UltralyticsDetectorProvider
MMDetDetectorProvider is optional now.
bunch of SEGS manipulation nodes
SEGS Filter, SEGS Detailer, SEGS Paste, SEGS Preview, ...
ImpactWildcardsProcessor, ImpactWildcardsEncode
ReencodeLatent
And much more improvements....
v2.7.5
Bug fixes: EmptySEGS, device error on Mask operation nodes,
improve install script for better compatibility with portable version
v2.7.1
Bug fix in the 'MASK to SEGS' node where an erroneous SEGS was generated when the crop region extended beyond the image area.
v2.7
Whole new nodes and New install system
Interactive SAM Detector, PreviewBridge
Iterative Upscaler, PIPE, ...
FaceDetailer
v1.7
new features: mask-points, mask-area on SAMDetectorCombined
improve mask painting: wheel brush size, right button eraser
SAMDetectorCombined bbox range bug fix
v1.6.1
bug fix for DetailerForEachDebug
v1.6
MaskPainter, EmptySEGS node added
bug fix for MaskToSEGS
v1.5
guide_size_for parameter added
more predictable upscale size patch
crop_factor on ONNXDetectorForEach node
v1.4
guide_size bug fix
ONNXLoader, ONNXDetectorForEach nodes added
v1.3
MaskToSEGS node added.
v1.2
Support external_seed for Seed node of WAS node suite.
v1.1
Fixed a package dependency issue with pycocotools on Windows.
Resolved an issue where the software was unable to recognize the "ComfyUI" folder in certain cases.
Description
FAQ
Comments (25)
"External seed" seems an extra parameter. Normal seed can be converted and connected to a primitive.
I implemented this feature because it was requested by someone who needed the WAS seed node, but I am now questioning its necessity since external parameters are already supported as a built-in feature in ComfyUI.
@DrLtData i'm really sorry, it could be me ;-)
This looks incredible. Thank you for sharing. I'm excited to learn about the tool you created
Is there any possibility to auto detect a mask based on a test input, eg "arms", "shoe" or whatever?
Based on a custom node in CLIPSeg, you can create a mask of a specific part using a prompt. This can then be converted to MaskToSEGS for further processing. The combination is so good.
I recommend this custom node.
https://github.com/biegert/ComfyUI-CLIPSeg
@DrLtData Oh wow...I didn't even think this was possible without training a custom Detection model. Thank you so much.
I'm a little confused about that, can you upload the full node image?
https://github.com/ltdrdata/ComfyUI-Impact-Pack
Here is a detailed tutorial. The current version of Civitai is an old version, and an update is scheduled once sufficient testing of the new version has been completed.
Hello! Please, could someone help me with this error I get with the first run?
ERROR: Could not install packages due to an OSError: [WinError 5] Acces denied: 'D:\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\cv2\\cv2.pyd'
Consider using the --user option or check the permissions.
Thank you!
The "--user" option cannot be applied when using a venv environment, so it was not being used. Instead, I have made a modification to the code so that opencv-python is installed only if cv2 is not present, and it is now installed together with other requirements. Please verify if the issue still persists.
@DrLtData Hi, thank you. I tried 'git pull' to obtain a new version. Some dependencies installed correctly, but at he end, this same error persist (I think):
Installing collected packages: yapf, opencv-python, mmengine, mmcv
ERROR: Could not install packages due to an OSError: [WinError 5] Přístup byl odepřen: 'D:\\ComfyUI_windows_portable\\python_embeded\\Lib\\site-packages\\cv2\\cv2.pyd'
Consider using the --user option or check the permissions.
Traceback (most recent call last):
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\install.py", line 76, in ensure_mmdet_package
import mmcv
ModuleNotFoundError: No module named 'mmcv'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "D:\ComfyUI_windows_portable\ComfyUI\nodes.py", line 1304, in load_custom_node
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in callwith_frames_removed
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\__init__.py", line 16, in <module>
import install # to install dependencies
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\install.py", line 117, in <module>
install()
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\install.py", line 89, in install
ensure_mmdet_package()
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\install.py", line 81, in ensure_mmdet_package
subprocess.check_call([sys.executable, '-m', 'mim', 'install', 'mmcv==2.0.0'])
File "subprocess.py", line 369, in check_call
subprocess.CalledProcessError: Command '['D:\\ComfyUI_windows_portable\\python_embeded\\python.exe', '-m', 'mim', 'install', 'mmcv==2.0.0']' returned non-zero exit status 1.
Cannot import D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack module for custom nodes: Command '['D:\\ComfyUI_windows_portable\\python_embeded\\python.exe', '-m', 'mim', 'install', 'mmcv==2.0.0']' returned non-zero exit status 1.
Since I really don't understand a word ;-) I maybe ask a stupid question. Might it be, because I use CPU only?
@Stepaan It's not related to whether CPU mode, it seems related to specific settings with portable and windows. I think I need to modify my install script. I'll let you know after fix.
I need some experiment.
I added '--user' option if comfyui is running on portable version. Try updated version.
@DrLtData Thank you so much. It worked like a charm (downloaded and installed a bunch of files), until:
### ComfyUI-Impact-Pack: Check basic models
Traceback (most recent call last):
File "D:\ComfyUI_windows_portable\ComfyUI\nodes.py", line 1304, in load_custom_node
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in callwith_frames_removed
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\__init__.py", line 37, in <module>
import impact_server # to load server api
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\impact_server.py", line 8, in <module>
import impact_core as core
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\impact_core.py", line 3, in <module>
import mmcv
ModuleNotFoundError: No module named 'mmcv'
Cannot import D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack module for custom nodes: No module named 'mmcv'
:-(
@Stepaan mmcv should installed before that point.
Try this at "D:\ComfyUI_windows_portable" in cmd.
.\python_embeded\python.exe -m mim install mmcv==2.0.0
@DrLtData Strange. Got it:
Requirement already satisfied: mmcv==2.0.0 in c:\users\union\appdata\roaming\python\python310\site-packages (2.0.0) ...etc, etc.
But later:
### ComfyUI-Impact-Pack: Check basic models
Traceback (most recent call last):
File "D:\ComfyUI_windows_portable\ComfyUI\nodes.py", line 1304, in load_custom_node
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in callwith_frames_removed
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\__init__.py", line 37, in <module>
import impact_server # to load server api
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\impact_server.py", line 8, in <module>
import impact_core as core
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\impact_core.py", line 3, in <module>
import mmcv
ModuleNotFoundError: No module named 'mmcv'
Cannot import D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack module for custom nodes: No module named 'mmcv'
@Stepaan Try updated version. Now install script will ignore global packages forcefully.
@DrLtData you're quick as a fox 😂 I'm going to try it in an hour or so. Thank you so much! Does it mean, I have two versions of it (one in the ComfyUI portable directory and another one in my python installation from A1111) ? And they're fighting each other?
@DrLtData Still the same. Differrent lines, but same errors. Nevermind, I don't want to steal your time. It seems, it's something related to my machine.
At the very beginning:
### Loading: ComfyUI-Impact-Pack (V2.7.7)
### ComfyUI-Impact-Pack: Reinstall dependencies (several dependencies are missing.)
### ComfyUI-Impact-Pack: Check dependencies
.
.
then later:
Using cached https://download.openmmlab.com/mmcv/dist/cu117/torch2.0.0/mmcv-2.0.0-cp310-cp310-win_amd64.whl (35.3 MB)
Requirement already satisfied: addict in d:\comfyui_windows_portable\python_embeded\lib\site-packages (from mmcv==2.0.0) (2.4.0)
.
.
and:
Requirement already satisfied: python-dateutil>=2.7 in d:\comfyui_windows_portable\python_embeded\lib\site-packages (from matplotlib->mmengine>=0.2.0->mmcv==2.0.0) (2.8.2)
[several times in several ocasions]
.
.
and finally:
Traceback (most recent call last):
File "D:\ComfyUI_windows_portable\ComfyUI\nodes.py", line 1354, in load_custom_node
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in callwith_frames_removed
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\__init__.py", line 37, in <module>
import impact_server # to load server api
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\impact_server.py", line 8, in <module>
import impact_core as core
File "D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\impact_core.py", line 3, in <module>
import mmcv
ModuleNotFoundError: No module named 'mmcv'
Cannot import D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack module for custom nodes: No module named 'mmcv'
:-(
@Stepaan
If the external package dependencies continue to cause issues, it may be necessary to try creating a venv (virtual environment) and use it to isolate the dependencies.
Can you setup venv of python?
@DrLtData I have no idea what that is. But I'll Google it. Thank you.
I'm also having the same problem as @Stepaan, I am running linux mint (ubuntu 22.04) when I try to install mmcv I get an error that I don't have torch. I'm running it in a conda env. Should I set it up in a python env? I'll try it out to see if it falls apart and get back here with some of the steps that I have to take to get it working.
@DrLtData @Stepaan Here is the solution my conda env was getting in the way and I needed to make a python env for this to work. Here are the steps.
I did this in my comfy dir
#make env
python -m venv venv
#activate venv
source venv/bin/activate
#install torch and deps that you need for comfy nvidia card for me
pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu118 xformers
#install the extra requirements
pip install -r requirements.txt
#everything should work now, you can run your comfy install with the node working in the venv
python main.py
I had some other deps that i needed for the was node set as well. Had to install numba but everything is showing as working now. I hope this helps the next person wanting to get comfy.