marascott / comfyui_marascott_nodes Goto Github PK

View Code? Open in Web Editor NEW

72.0 72.0 9.0 28.43 MB

A set of community node including a universal bus and a large Upscaler/Refiner

Home Page: https://www.marascott.ai/

License: Other

Python 73.97% JavaScript 26.03%

bus comfyui comfyui-nodes refiner upscaler

comfyui_marascott_nodes's People

Contributors

Stargazers

Watchers

Forkers

navezjt luxcium gonzalu haohaocreates jeyamir jags111

comfyui_marascott_nodes's Issues

McBoaty : deepshrink and co idea

build in DeepShrink or similar functionality or use StableCascade to generate natively above the magical 1024...and SAG or Perturbed whatever and of course CONTROLNets to tighten the generation of the upscales

Yeah but there is also deepshrink and Cascade where you basically tuned the compressor directly in the Latent generator... I have used the max compression (I think it's 64) and generated 2048x2048 without any issues.

Cascade also does better text.

AnyBus : Create Predefined Profile

Allow user to select predefined profile :

basic_pipe
detailer_pipe
Custom (default behavior)
User_profiles

User can register profiles which will then be saved and listed in the profile list

Custom Profile is the default Behavior

Implement Mickey Resize to Nearest SDXL

check Mickey Resize to Nearest SDXL node to preconfigure input image to SDXL sizes chosen by a select

McBoaty - implement VRAM debug from KJNodes to release resources

https://github.com/kijai/ComfyUI-KJNodes

McBoaty - VAE Encode/Decode Tiled

Use VAE Encode and Decode Tiled when necessary

McBoaty : add center tile preview test

add a feature to test center tile and do not run the entire flow

check UE works with BUS Input

BusNode do not add

check the cache

add Detailer pipe bus

Controlnet Support or IP Adapter Tiles Support

Does the McBoaty Upscaler work if you string controlnet or Ip Adapter into it?

Implement profile

set a widget to define a name which will trigger the sync only for nodes with same profile name
by default a node is set to 'default'

AnyBus : Set/Get

When using the GetSet node we need to assign the same profile name to the setter but if you need to chain it, you can't

Integrate an SetGet node system to handle automaticaly the profile name

or use ComfyUI-mape-Helpers wireless node which will be nicer
all support for MapeVariable

allow to keep entries set on bus disconnect

link is passed to null on page reload

c2fe112

here is the file diff after a page reload

@@ -327,7 +327,7 @@
         {
           "name": "bus",
           "type": "BUS",
-          "link": 401
+          "link": null
         },
        {
          "name": "* 01 - model",
@@ -841,7 +841,7 @@
        {
          "name": "bus",
          "type": "BUS",
-          "link": 403
+          "link": null
          "link": null
        },
        {
          "name": "* 01 - model",

does not support tiled controlnet when upscale image

I will apply controlnet with original image to each tile, and end up with something like this. It does not work even if you set the output size as same size.

Images won't pipe, throws tuple error.

Failed to validate prompt for output 2621:

(prompt):
- Exception when validating node: tuple index out of range
PreviewImage 2621:
- Exception when validating node: tuple index out of range
  Output will be ignored
  Failed to validate prompt for output 2584:
(prompt):
- Required input is missing: images
Image Save 2584:
- Required input is missing: images
  Output will be ignored

McBoaty : apply mask to fix seams

https://discord.com/channels/1197409727521509506/1212901064655769610/1237851947395186809

'No module named 'comfy_extras.nodes_align_your_steps'

ComfyUI on mac,

I tried to install using comfyUI manager. but after installing, I get this error :
Cannot import /Users/lalala/pinokio/api/comfyui.git/app/custom_nodes/ComfyUI_MaraScott_Nodes module for custom nodes: No module named 'comfy_extras.nodes_align_your_steps'

I googled this module with no success...

Any idea ?

Thanks

multi shared flows do not update properly the frontend data

when generating 2 flows starting with the same AnyBus it works but the display of profile and inputs names on nodes is not populated correctly

MultiBusNodeMerging ?

Allow to manage inputs/outputs label per flows

If i feed "Load Image" boxes it displays just image on each, hard to figure out wich is wich. Grabing the Title of the connected node would be nice.

Error no module named blend_modes

Hi,
It seems that this package takes precedence over was-node-suite-comfyui when installing dependencies with comfy-cli:
comfy node install-deps --workflow <myworkflow.json> will now install ComfyUI_MaraScott_Nodes rather than the WAS package because this package includes several others.

The error encountered is:

Traceback (most recent call last):
 File "/home/user/repos/ComfyUI/nodes.py", line 1879, 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 _call_with_frames_removed
 File "/home/user/repos/ComfyUI/custom_nodes/ComfyUI_MaraScott_Nodes/__init__.py", line 55, in <module>
   from .MaraScott_Nodes import NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS, WEB_DIRECTORY
 File "/home/user/repos/ComfyUI/custom_nodes/ComfyUI_MaraScott_Nodes/MaraScott_Nodes.py", line 17, in <module>
   from .py.nodes.KSampler.InpaintingTileByMask_v1 import KSampler_setInpaintingTileByMask_v1, KSampler_pasteInpaintingTileByMask_v1
 File "/home/user/repos/ComfyUI/custom_nodes/ComfyUI_MaraScott_Nodes/py/nodes/KSampler/InpaintingTileByMask_v1.py", line 19, in <module>
   from ...vendor.ComfyUI_LayerStyle.py.image_blend_v2 import ImageBlendV2, chop_mode_v2
 File "/home/user/repos/ComfyUI/custom_nodes/ComfyUI_MaraScott_Nodes/py/vendor/ComfyUI_LayerStyle/py/image_blend_v2.py", line 1, in <module>
   from .imagefunc import torch, Image, tensor2pil, pil2tensor, chop_mode_v2, chop_image_v2
 File "/home/user/repos/ComfyUI/custom_nodes/ComfyUI_MaraScott_Nodes/py/vendor/ComfyUI_LayerStyle/py/imagefunc.py", line 40, in <module>
   from .blendmodes import *
 File "/home/user/repos/ComfyUI/custom_nodes/ComfyUI_MaraScott_Nodes/py/vendor/ComfyUI_LayerStyle/py/blendmodes.py", line 10, in <module>
   from blend_modes import difference, normal, screen, soft_light, lighten_only, dodge,   \
ModuleNotFoundError: No module named 'blend_modes'

Cannot import /home/user/repos/ComfyUI/custom_nodes/ComfyUI_MaraScott_Nodes module for custom nodes: No module named 'blend_modes

blend_modes is one of the dependencies of ComfyUI_LayerStyle which is not properly installed during install process when pulling the node through the cm-cli/comfy-cli

Although this could be a question for the comfy-cli or manager repos, do you guys know a way to not use the ComfyUI_MaraScott_Nodes ? It is a really nice node package but I feel like it is way too big for my needs and I would like to be able to use the was_node_suite.
Also I am curious about what drove the decision to package third party node packages in this one instead of considering them as an external dependency.

Thank you !

Add data to generated image while doing an XYplot

Request :

https://discord.com/channels/1197409727521509506/1200021802689253427/1216148814402945145
https://discord.com/channels/1197409727521509506/1200021802689253427/1216181388789813298
https://discord.com/channels/1197409727521509506/1200021802689253427/1216304805681827932

Target workflow : Video_Generation_Workflow.json

AnyBus : Input Combo & primitive

Why Combo and primitive can t be pluged in ?

McBoaty - check sigma

Check sigma code and custom ksampler

McBoaty: option to ByPass the UpscaleByModel

on/off option with a x1.2 (?) upscale then back to 1:1 resolution

Implement Get/Set system

allow use of get / set node for Bus
allow get to bus only if origin set is Bus node from our node
allow connection from bus only for set node or our node

Node request : Prompt Morphing

https://www.youtube.com/watch?v=Aw2Z27A0vVg

rename entries

Complex workflow to test

https://discord.com/channels/752483298198028309/1084878026644393984/1207794672651730995

https://discord.com/channels/752483298198028309/1084878026644393984/1207856021234647101

Error executing Large Refiner McBoaty v2

I am getting this error on a fairly fresh install of ComfyUI.

!!! Exception during processing!!! module 'comfy_extras' has no attribute 'nodes_upscale_model'
Traceback (most recent call last):
  File "E:\dev\ComfyUI_portable\ComfyUI\execution.py", line 151, in recursive_execute
    output_data, output_ui = get_output_data(obj, input_data_all)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "E:\dev\ComfyUI_portable\ComfyUI\execution.py", line 81, in get_output_data
    return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "E:\dev\ComfyUI_portable\ComfyUI\execution.py", line 74, in map_node_over_list
    results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "E:\dev\ComfyUI_portable\ComfyUI\custom_nodes\ComfyUI_MaraScott_Nodes\py\nodes\UpscalerRefiner\McBoaty_v2.py", line 167, in fn
    upscale_model = comfy_extras.nodes_upscale_model.UpscaleModelLoader.load_model(comfy_extras.nodes_upscale_model.UpscaleModelLoader, upscale_model_name)[0]
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: module 'comfy_extras' has no attribute 'nodes_upscale_model'

I have the following custom nodes installed.

Here is my console prior to running the workflow

## ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2024-05-14 14:06:32.826101
** Platform: Windows
** Python version: 3.11.8 (tags/v3.11.8:db85d51, Feb  6 2024, 22:03:32) [MSC v.1937 64 bit (AMD64)]
** Python executable: E:\dev\ComfyUI_portable\python_embeded\python.exe
** Log path: E:\dev\ComfyUI_portable\comfyui.log
[Maras IT] Initialization

Prestartup times for custom nodes:
   0.0 seconds: E:\dev\ComfyUI_portable\ComfyUI\custom_nodes\ComfyUI_MaraScott_Nodes
   0.2 seconds: E:\dev\ComfyUI_portable\ComfyUI\custom_nodes\ComfyUI-Manager

Total VRAM 24563 MB, total RAM 65478 MB
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 4090 : cudaMallocAsync
VAE dtype: torch.bfloat16
Using pytorch cross attention
### Loading: ComfyUI-Manager (V2.33)
### ComfyUI Revision: 2180 [2de3b69b] | Released on '2024-05-13'
### Loading: Save img prompt
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
------------------------------------------
Comfyroll Studio v1.76 :  175 Nodes Loaded
------------------------------------------
** For changes, please see patch notes at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/blob/main/Patch_Notes.md
** For help, please see the wiki at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/wiki
------------------------------------------
### [START] ComfyUI AlekPet Nodes ###
Node -> ArgosTranslateNode [Loading]
Node -> DeepTranslatorNode [Loading]
Node -> ExtrasNode [Loading]
Node -> GoogleTranslateNode [Loading]
Node -> PainterNode [Loading]
Node -> PoseNode [Loading]
### [END] ComfyUI AlekPet Nodes ###
[Maras IT] Loaded
[comfy_mtb] | INFO -> loaded 73 nodes successfuly
[comfy_mtb] | INFO -> Some nodes (5) could not be loaded. This can be ignored, but go to http://127.0.0.1:9188/mtb if you want more information.

Import times for custom nodes:
   0.0 seconds: E:\dev\ComfyUI_portable\ComfyUI\custom_nodes\websocket_image_save.py
   0.0 seconds: E:\dev\ComfyUI_portable\ComfyUI\custom_nodes\ComfyUI-SaveImgPrompt
   0.0 seconds: E:\dev\ComfyUI_portable\ComfyUI\custom_nodes\ComfyUI_YFG_Comical
   0.1 seconds: E:\dev\ComfyUI_portable\ComfyUI\custom_nodes\ComfyUI_MaraScott_Nodes
   0.2 seconds: E:\dev\ComfyUI_portable\ComfyUI\custom_nodes\ComfyUI-Manager
   0.3 seconds: E:\dev\ComfyUI_portable\ComfyUI\custom_nodes\ComfyUI_Comfyroll_CustomNodes
   0.7 seconds: E:\dev\ComfyUI_portable\ComfyUI\custom_nodes\comfy_mtb
   5.0 seconds: E:\dev\ComfyUI_portable\ComfyUI\custom_nodes\ComfyUI_Custom_Nodes_AlekPet

Starting server

To see the GUI go to: http://127.0.0.1:9188

and here is the workflow

Error_workflow.json

implement impack control bridge blocking system

Add AlignYourStepsScheduler to McBoaty

from ComfyUI\comfy_extras\nodes_align_your_steps.py

AlignYourStepsScheduler.get_sigmas(self, model_type, steps, denoise)

model_type = ["SD1", "SDXL", "SVD"]

comfy_extras.nodes_align_your_steps.AlignYourStepsScheduler.INPUT_TYPES().required.model_type

check old workflow with new node

https://discord.com/channels/@me/1199990759408025611/1208135246285897728

AnyBus : cleaning

find a way to clean the inputs

McBoaty : profiles

like profile for settings A and one for B etc. So I can load presets of combos of Sampler/Checkpoint/Steps/Denoise/UpscaleModel/etc.

Node Cleaning when no input bus is connected

reset input/output

apply color to flow

Implement add/remove input/output type ANY

Add a menu item "Add Input" and "Remove Input" similar to MultiAreaConditioning Node

McBoaty : LLM

Get inspiration from Plushs

https://github.com/glibsonoran/Plush-for-ComfyUI

Add documentation helper

check KJnodes addDocumentation in JS

AnyBus : dashboard

find a way to centralize the node profiles

InpaintingByMask

suggestion to grow initial mask area

McBoaty : Check Was Ksampler

WAS cyclic ksampler

Check if conditioning is passed by AnyBus properly

Weird, its seems like when using the any bus, the conditioning doesn't get passed through. First image (the lady one), the conditioning is ignored when coming through the anybus & the second is using the exact same text directly. Anyway, probably not important at this stage. (this is ofc using that extension called out)

McBoaty : memory release

Right I hit a serious glitch. Python is not releasing memory between generations, so do a few in a row and the whole system goes down. The vae stuff and the node data needs to be cached to disk not held in system memory.

WAS Node Suite

inspire has some cache nodes, for once

Issue

when i hit queue, i get a out of memory error. then i hit just again queue, and it de-caches all kinds of stuff from before and the workflow goes on.

Image processing is trigger before the workflow is actualy executed

when I execute, the mtb debug node will receive the image information on due time but the previewImage node "receive" the image information at the beginning of the execution before it is processed

if I link the image output from vae decode directly to preview image, I have the same effect from the VAE decode node, it is "executed before everything else

but the mtb debug is still processed in due time

Here the python console log when running the workflow

got prompt
ERROR:root:Failed to validate prompt for output 78:
ERROR:root:* MaskDetailerPipe 75:
ERROR:root:  - Exception when validating inner node: tuple index out of range
ERROR:root:Output will be ignored
ERROR:root:Failed to validate prompt for output 24:
ERROR:root:* (prompt):
ERROR:root:  - Exception when validating node: tuple index out of range
ERROR:root:* PreviewImage 24:
ERROR:root:  - Exception when validating node: tuple index out of range
ERROR:root:Output will be ignored
ERROR:root:Failed to validate prompt for output 71:
ERROR:root:* (prompt):
ERROR:root:  - Exception when validating node: tuple index out of range
ERROR:root:* PreviewImage 71:
ERROR:root:  - Exception when validating node: tuple index out of range
ERROR:root:Output will be ignored
ERROR:root:Failed to validate prompt for output 80:
ERROR:root:* DetailerForEachPipe 79:
ERROR:root:  - Exception when validating inner node: tuple index out of range
ERROR:root:Output will be ignored
[rgthree] Using rgthree's optimized recursive execution.
[rgthree] First run patching recursive_output_delete_if_changed and recursive_will_execute.
[rgthree] Note: If execution seems broken due to forward ComfyUI changes, you can disable the optimization from rgthree settings in ComfyUImodel_type EPS
adm 2816
Using pytorch attention in VAE
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
Using pytorch attention in VAE
missing {'cond_stage_model.clip_l.logit_scale', 'cond_stage_model.clip_l.text_projection'}
left over keys: dict_keys(['cond_stage_model.clip_l.transformer.text_model.embeddings.position_ids'])
Requested to load SDXLClipModel
Loading 1 new model
Requested to load SDXL
Loading 1 new model
100%|███████████████████████████████████████████████████████████████████████████| 10/10 [00:03<00:00,  2.55it/s]
Requested to load AutoencoderKL
Loading 1 new model
Prompt executed in 13.84 seconds

McBoaty Upscaler - ability to process individual tiles

Thanks a lot for converting into a node!

Feature request: Can a user chain these nodes such that Node A passes the tiles from its run into Node B and then is able to run B to target a specific tile with different denoise, feather settings? (~9m into Rob Adams video for details).