AP Plot 1.0 for ComfyUI
In 2023, Alessandro created the AP Workflow for ComfyUI, an advanced automation workflow to orchestrate Stable Diffusion and dozens of other AI models.
Since then, AP Workflow has been downloaded over 30,000 times by inviduals and organizations all around the world.
AP Plot (APP) is Alessandro’s second workflow for ComfyUI, designed to solve a specific problem: evaluating and comparing the best Stable Diffusion fine-tuned models for photorealism.
AP Plot 1.0 focuses SDXL fine-tunes, but it can be reconfigured to compare anything you like.
The list of models being compared in this first release:
- STOIQO NewReality XL 4.0
- ICBINP XL 6
- epiCRealism XL 7
- RealVis XL 4.0
- HelloWorld XL 7.0
- Juggernaut X XL
To compare the behavior of these modes with a wide range of prompts, Stable Diffusion users would typically use an XY Plot.
The problem is that in a comparison with a traditional XY Plot (including the one in AP Workflow), each model is used with the same set of parameters.
That might seem a fairer comparison, but it doesn’t help you see the true potential of each model.
AP Plot allows you to have the best of both worlds, and then some:
- It compares the output of each model according to the parameters indicated by each model author.
- It compares the output of each model according to the parameters Alessandro personally recommends, identical for each model.
- It compares the output of each model according to parameters defined by you, identical for each model.
In other words, AP Plot can show you an apple-to-orange comparison side by side with an apple-to-apple comparison. And it comes all preconfigured, so you don’t have to worry about much except downloading the models.
And just like AP Workflow, AP Plot is free.
Keep reading and download AP Plot 1.0 for ComfyUI.
Content
Jumpstart
Support
Extras
Jumpstart
How to Download AP Plot 1.0
Download the JSON version of AP Plot 1.0 and load it via ComfyUI Manager. That’s it!
AP Plot 1.0 is also embedded in the workflow picture below. Click on it, and the full version will open in a new tab. Right click on the full version image and download it. Drag it inside ComfyUI, and you’ll have the same workflow you see below.
Be sure to review the Required Custom Nodes and Required AI Models sections of this document before you download APW 10.0.
Early Access Program
AP Workflow 11.0 EA5
AP Workflow for ComfyUI early access features available now:
- [EA5] The Discord Bot function is now the Bot function, as AP Workflow 11 now can serve images via either a Discord or a Telegram bot.
- [EA5] The Bot function now can send a helpful message to the user with the list of commands that can be used.
- [EA5] AP Workflow now automatically saves an extra image in JPG and stripped of all metadata, for peace-of-mind sharing on social media.
- [EA5] The Uploader function now allows you to iteratively load the 1st reference image from a source folder. This is particularly useful to process a large number of images without the limitations of a batch.
- [EA4] AP Workflow now features two extra functions: FLUX ControlNet Tile and FLUX ControlNet Pose, enabled by the InstantX ControlNet Union Pro model.
- [EA4] The Inpainter and Repainter (img2img) functions now use FLUX as default model.
- [EA4] The Discord Bot function now allows to set a third style for experimentation. You can invoke it with the flag --s test.
- [EA3] The Discord Bot function now allows to customize a series of sentences that the bot returns to the user to acknowledge the generation request.
- [EA3] The Discord Bot function now supports user requests for photographic (default), artistic, and clay image style.
- [EA2] AP Workflow now supports the new ControlNet Canny, Depth, and HED models for FLUX.
- [EA2] AP Workflow now supports FLUX Model Sampling.
- [EA2] The Discord Bot function now supports user requests for landscape (default), portrait, square, and cinema image format, as well as custom sizes.
- [EA2] The Discord Bot function now returns the full list of parameters used to generate an image.
- [EA2] AP Workflow features a new LUT Applier function, useful to apply a Look Up Table (LUT) to an uploaded or generated image.
- [EA2] AP Workflow features a new Color Corrector function, useful to modify gamma, contrast, exposure, hue, saturation, etc. of an uploaded or generated image.
- [EA2] AP Workflow features a new Grain Maker function, useful to apply film grain to an uploaded or generated image.
- [EA1] AP Workflow is completely redesigned. Too many changes to list them all.
- [EA1] AP Workflow now supports the new FLUX 1 model and its LoRAs.
- [EA1] AP Workflow now features five independently-configured pipelines, so you don't have to constantly reconfigure the parameters:
- Stable Diffusion 1.5 / SDXL
- FLUX 1
- Stable Diffusion 3
- Dall-E 3
- Repainter (img2img with SDXL)
- [EA1] The ControlNet functions (Tile, Canny, Depth, OpenPose) now feature the new ControlNet SDXL Union Promax. As before, each function can be reconfigured to use a different ControlNet model and a different pre-processor.
- [EA1] AP Workflow now features the capability to receive commands and return images (and text) via your own Discord bot.
- [EA1] AP Workflow now features a brand new, highly granular and customizable logging system.
- [EA1] The Upscaler (SUPIR) function now automatically generates a caption for the source image to upscale via a dedicated Florence-2 node.
- [EA1] The Inpainter without Mask function is now called Repainter (img2img).
AP Plot 2.0 EA2
AP Plot for ComfyUI early access features available now:
- [EA2] AP Plot now comes with 14 prompts to use for comparisons.
- [EA2] Approach 3 is now dedicated to testing each chosen model with the addition of a LoRA.
- [EA2] Approach 2 now features a Sampler Custom node instead of the default KSampler to accommodate changes in the Fixed Generation Parameters function.
- [EA1] AP Plot replaced HelloWorld XL 7.0 with the new FLUX 1 Dev.
How to Join
Get an Early Access membership via Patreon or Ko-fi, and you'll have access to a dedicated Discord server where Alessandro shares the unfinished new versions of AP Workflow as he progresses in its development.
Pros
- You'll gain a competitive edge at work!
- You'll be able to provide early feedback on the AP Workflow design and potentially influence its development.
- You'll support the future development of the AP Workflow.
- There will be no documentation.
- Things might change without notice.
- There is no guarantee you'll receive support.
Stay In The Loop
Where to Start?
AP Plot features a modular design. There’s no wire going from one group to another. Which means that you can move groups around to completely change the design of AP Plot.
Four areas are dedicated to the workflow configuration, three areas are dedicated to image generation, and three areas are dedicated to image inspection and saving.
Workflow Configuration
0 – The control panel where you centrally enable or disable all the other groups.
1 – The list of models being compared.
2 – The prompt builder.
As AP Plot is not meant to replace AP Workflow in any way, this is very simple and only allows you to choose between a few alternatives.
3 – The configuration parameters Alessandro recommends.
These should not be touched as they are only used by group #5.
Image Generation
4 – Image generation according to the parameters indicated by each model author (Approach 1).
These recommendations are retrieved from the Civitai pages indicated in the Notes. Occasionally, the recommendations are not crystal clear and I had to guess a bit.
This is NOT an apple-to-apple comparison.
5 – Image generation according to the parameters I recommend, identical for each model (Approach 2).
These parameters are defined in group #3.
This is an apple-to-apple comparison, even if an unfair one.
6 – Image generation according to parameters defined by you (Approach 3).
Customize this area for experimentation as you like.
Image Inspection and Saving
7 – Image comparison for different approaches for each tested mode.
By default, each node is configured to compare Approach 1 and Approach 2, but you can modify it to compare 1 and 3, or 2 and 3.
8 – Individual image saver.
You can disable this group if you just want to have a composite image, but this is where you do your main inspection.
9 – Composite image saver.
It arranges individual images in a grid and adds a small label in the top left corner.
At the moment, I cannot create a single image with all 18 images, so this is split into two composite images with 9 images each. In the future, there will be a single grid.
Support
Required Custom Nodes
If, after loading the workflow, you see a lot of red boxes, you must install some custom node suites.
AP Plot depends on multiple custom nodes that you might not have installed. You should download and install ComfyUI Manager, and then install the required custom nodes suites to be sure you can run this workflow.
You have two options:
- Recommended: Restore AP Workflow custom node suites snapshot.
- Discouraged: Install the required custom node suites manually.
Restore AP Workflow custom node suites snapshot
The manual installation of the required custom node suites has been proven unreliable for many users. To improve your chances to run this workflow, you should create a clean install of ComfyUI, manually install the ComfyUI Manager, and then restore a snapshot of Alessandro’s ComfyUI working environment.
This will help you install a version of each custom node suite that is known to work with AP Plot 1.0.
Instructions:
- Install ComfyUI in a new folder to create a clean, new environment (a Python 3.10 venv is recommended).
- Install ComfyUI Manager.
- Shut down ComfyUI.
- Download the snapshot.
- Move/copy the snapshot to the /ComfyUI/custom_nodes/ComfyUI-Manager/snapshots folder.
- Restart ComfyUI.
- Open ComfyUI Manager and then the new Snapshot Manager.
- Restore the AP Workflow 10.0 Custom Nodes Snapshot.
- Restart ComfyUI.
Notice that the Snapshot Manager is a relatively new, experimental feature and it might not work in every situation. If you encounter errors, check the documentation: https://github.com/ltdrdata/ComfyUI-Manager#snapshot-manager
Install the required custom node suites manually
Install the nodes listed below via ComfyUI Manager.
Notice that some of these nodes conflict with others that you might have already installed. This is why this option is highly discouraged and you should not use it unless you know how to troubleshoot a ComfyUI environment.
- \ComfyUI\custom_nodes\ComfyUI-Custom-Scripts
- \ComfyUI\custom_nodes\ComfyUI-KJNodes
- \ComfyUI\custom_nodes\rgthree-comfy
- \ComfyUI\custom_nodes\comfyui-prompt-reader-node
- \ComfyUI\custom_nodes\ComfyUI-Manager
- \ComfyUI\custom_nodes\ComfyUI-Impact-Pack
- \ComfyUI\custom_nodes\was-node-suite-comfyui
- \ComfyUI\custom_nodes\comfy_mtb
AttributeError: module ‘cv2.gapi.wip.draw’ has no attribute ‘Text’
after which, ComfyUI will fail to import multiple custom node suites at startup.
If so, you must perform the following steps:
- Terminate ComfyUI.
- Manually enter its virtual environment from the ComfyUI installation folder by typing: source venv/bin/activate
- Type: pip uninstall -y opencv-python opencv-contrib-python opencv-python-headless
- Type: pip install opencv-python==4.7.0.72
- Restart ComfyUI.
Required AI Models
The SDXL fine-tuned models compared in AP Plot 1.0 can be found on the Civitai website or on the Hugging Face website.
AP Plot only compares models that are publicly available and free to access.
Should a model become paywalled, it will be removed from AP Plot.
Node XYZ failed to install or import
Occasionally, AP Workflow users can’t install or import a custom node suite necessary to run AP Workflow. This happens when you try to use AP Workflow in a pre-existing ComfyUI environment that you have installed loooooooooong time ago.
If you have a similar problem, be sure to:
-
Have all your packages up to date in your Python virtual environment for ComfyUI.
To do so:- Terminate ComfyUI.
- Manually activate its Python virtual environment with source /venv/bin/activate.
- Run pip install -U pip to upgrade pip.
- Run pip install -U setuptools to upgrade setuptools.
- Run pip install -U wheel to upgrade wheel.
- Run pip install -U -r requirements.txt to upgrade all the packages in the virtual environment.
- Check the installation instructions of the custom node suite you are having troubles with.
If you have installed ComfyUI in a new environment and you still fail to install or import a custom node suite, open an issue on the GitHub repository of the author.
I Need More Help!
AP Workflow and AP Plot are provided as is, for research and education purposes only.
However, if your company wants to build commercial solutions on top of ComfyUI and you need help with these workflows, you could work with Alessandro on your specific challenge.
Extras
Special Thanks
AP Workflow wouldn’t exist without the dozen of custom nodes created by very generous members of the AI community.
In particular, special thanks to:
@rgthree:
- His Fast Groups Muters node powers AP Plot control panel.
- His Image Comparer node is critical to inspect and compare the behavior of the models in AP Plot.
@kijai:
- His Set and Get nodes enable the modular design of AP Plot.
- His ImageGridComposite node generates a convenient XY Plot image.
@receyuki:
- His SD Parameter Generator node is the most powerful and convenient way to define the Generation Parameters in AP Plot.
@ltdrdata:
- His ComfyUI Manager is critical to manage package and model dependencies in AP Plot.
Thanks to all of you, and to all other custom node creators for their help in debugging and enhancing their great nodes.
Full Changelog
- Initial Release.