Have you ever created an image you like and would have kept, but a face or hand just came out wrong? You need a workflow that can take control and address that.

Ever make an image that had multiple characters, but when you tried to assign them the correct eye color or expression, the checkpoint got confused, and you had to play generation roulette? You need a workflow that lets you decide whose face has to come out just so - no letting the model guess, total separation via individual detailers easily organized and there when you need them.

Testing out the new possibilities in Krea 2, and wanting to prompt more detail for things in the face and hands (or eyes, nsfw parts, anywhere that bboxes can be detected), but the granularity of detail is overwhelming your first pass prompt? Explore the possibilities of prompting individual SEGS boxes to add additional detail, using the power of this workflow.

I have been using this workflow for months now as my daily driver. I simply love it. I try other workflows but I come back to this one because I can rely on it. This ain't the kind of workflow that just falls out of a tree either. A lot of thought went into it. A lot of fiddling with booleans and logic gates. Before Krea 2, this workflow was my daily driver T2I workflow for checkpoints based in SDXL, Illustrious and Pony, and now Anima and Krea! It is based upon my Img2Img SDXL Workflow + META, which is where I first unleashed the power of the detailer array -- which revolutionized what was possible in SDXL at the time. I'm not saying I was the first, but I was doing it in the best way possible, which is with individually promptable SEGS bboxes and configurable SEGS sorting nodes which meant that there was no limit to how many boxes (theoretically) that I could prompt in one pass.

With version 7, each detailer type has three prompts instead of 5, which is still plenty. Big thanks to creators that helped me along the way, like Lonecatone23, Legendaer and Yukichan_IT.

Krea 2 Version 8.1

Krea v8.1 fixed some fairly serious bugs that snuck over with 8.0.1. I really apologize for this, I didn't notice it right away and perhaps I was a little too hasty with finalizing 8.0.1. As you can see from 8.0.1's metadata, I had been using the workflow just fine, but these problems somehow escaped my notice. Then it was a while before I could get back online. During that time, I noticed the problems, and was able to upload the fixes within 24 hours of noticing.. but that means, it was up for a few days all jacked up. But I fixed it asap: All this, for free, I do for you. Please thumbs up!

The bugs had to do with the way the prompts were sorted, to ensure that there was no doubling up of prompts. But I realized that if I rely on the LC Any Switch node, it will function properly and block the default positive prompt if any other prompts are there. So I simplified the switching and it did it properly.

The main difference with this version from version 7 is that there are now local AI bots you can query to either write a prompt for an image you give it, or write a prompt based on text you give it.

About Version 7 series:

SDXL 7.0 - Brand New, SDXL/Illustrious/Pony all work with this model

Krea 7.3 - Already had some bug fixes, make sure you are up to date.

Anima 7.3 - ensure you are setting CFG to 4.5-5 for non-turbo gens! Download Anima 7.0.1

These workflows are being published after fixing a bug that has been around for several versions now. It's complicated but basically the mechanism that I was using to control whether the detailer file was being saved was causing a second generation to run if the image saver node wasn't what was expected (i.e., right after toggling the detailer on, or off). I settled on a different set of nodes to accomplish this and couldn't be happier finally getting that logic puzzle solved. Wish I had tackled it sooner. Had to share my solution with everyone.

About Version 7.1

For those who downloaded 7.0 or 7.01, please update. Those versions had little bugs. The bypass core was checking too many things along the way and it was causing it to never pass the color-matched image along to downstream. This is pretty serious and caused me to need to issue a bug fix WITH a new version so everyone at least has a chance to see it. I knew about it and actually had it fixed since last night but posting it took time, I've been super busy, my house is still really hot too which makes testing risky business. I have found a workaround for my overheating PC though. Generate less, and generate smaller subject matter. These tiny birds were no trouble at all.

Seriously though, sorry. I really hate letting a mistake like that get past and then turn around and ask everyone to update. Hundreds downloaded my Krea workflows and you also came to try Lonecat's nodes so I really want to thank you for your patience. The problem with 7.0.1 was there was a missing boolean that had vanished while I was fixing the bypass core subgraph after 7.0 proved to have a bug. Without that boolean, the 3rd pass image save wasn't getting muted when it needed to, and throwing annoying errors because it was starved for an input. The bypass core due to a missing boolean link that was supposed to leave the bypass core, so now we're on the 3.1 bypass core. There's an annoying error where the bypass core does not send an image to the detailer saving node because the detailers are off, but a boolean connection to a muter was disconnected, causing the detailer saver node to be starved for inputs and throw errors instead of being muted.

About 7.0 (deprecated)

Building upon the improvements from 6.3, 7.0 comes with an extensive overhaul, new LC Nodes and a Post-Processing suite that puts the power to alter the image in amazing ways. You can even do a dynamic overlay in which the opacity of the final altered image is adjustable by you. Freeze the seed is the norm. The Input section also got an overhaul.

This has been a bit of a collaboration with @lonecatone23 so a big thanks to him and check out his workflows and models. Comfyui_LC123_nodes are listed on ComfyUI Manager so go get them, they're pretty much mandatory now or will be soon so just get them.

For tool designers, check out his example workflow contained within his custom nodes repo.

About Version 6.3

I made some small changes. Big shoutout to bkidderz for showing me how he does wildcards. In order to implement his methods, I had to change my wildcard processor. However, doing so didn't break anything including freezing the seed in order to freeze the wildcard results, a handy feature. All images made in the gallery utilized the new setup, and I often gained inspiration from the wildcards before finally settling on what I would actually depict. You might have much more success with your prompt, and remember that since Krea 2 is so good with multi-character prompts, you could wildcard that!

About Version 6.2

Made some small tweaks to the detailer arrays -- some detailers were not connected to all the primitives, leading to unnecessarily high number of steps, etc. (3rd and 4th detailers across the board). Also connected the denoise mask to the feather primitive, allowing you to change both at once.

The main change in 6.2 is to get LoRAs working again. I know right? Horrible. Read all about that disaster here.

About Version 6.0

This version came about due to my attempts to limit image-saving to not unnecessarily save extra copies of the upscaled image if the detailers were not also running. Now in theory I could further enhance the logic so that if the detailers didn't find any SEGS and thus didn't run, to not save the additional file, but the LOGIC for this, the way I did it, created an error, and I didn't know how to fix it! Thanks to Lonecatone23, who swooped in to save the day, patiently showing me how, you all can benefit and me too.

Lonecat (south of the border, this hombre is known as the Lone Catone) showed me how to make the workflow more efficient, saving time on upscales and removing redundant processes. After getting his fixes, I incorporated them into SDXL 6.0 with a full overhaul. Then I copied that workflow and made the adjustments needed to create an Anima T2I, SDXL I2I, and Anima I2I workflow. The I2I workflows contain the same resizer feature but I made it better and more robust, so now you can resize it or not and scale your resized image up (or down) prior to generating. Get set nodes have been added in some areas to reduce the stringiness of the workflow, although it remains fibrous in many areas. Of course, Metadata saving remains in force, making it a snap to share your images on this platform Enjoy!

Main Features:

✦ Text to Img Generation - That's what we're here for, text to image.

Include Metadata / Embed Workflow Option. Stable 1 megapixel generation, or dare to increase it with Size Boost.

✦ Metadata saving included — CivitAI ready. You can also embed the entire workflow into your images, by default this is turned off.

✦ Upscaler with Color Matching - this is the feature that makes images go from good to great. Or use HiresFix for a quick increase in resolution

✦ Quality Tags Presets Switch - instead of typing Masterpiece, best quality, etc. etc.

✦ Two Segs Detailer Arrays - two SEGS pipelines (by default, hands and faces) that you can individually prompt or apply one prompt to all. Sophisticated tool created to handle individual details - your backstop for when the image has to be perfect.

✦ Parallel development with I2I workflow - if you learn this one my Img2Img works essentially the same way.

✦ CFG Controls - need more control, more CFG? It's handled automatically, crank it up to 10.

✦ LoRA Manager — Search your LoRA database with style, I love this node. Dump it if you prefer something else.

✦ Simple and reliable — Enter your prompt, hit run, let the segs previews populate, freeze the gen seed, and fill in prompts and run it again. That's it.

New with 5.0 SE

Skipped 4.0, just because

Seed Jump

Style Preset switching

Added a toggle and Subgraph for exceeding native SDXL resolution on initial gen

Fixed Typos for name of Face Detailer LoRA blocker nodes

Does not include Quality Tags Fast Switching, uses native concatenate nodes

5.0.1 Changes:

Changed Positive Text Box to Mira, which was already being used in this workflow

Changed Wildcard Processor

Seed controls Wildcard selection (Freezing seed freezes your wildcards too now)

Added Final Prompt Preview

Added a toggle and Subgraph for exceeding native SDXL resolution on initial gen

Slight rearrangement of Control Group, hopefully optimal

New with 3.0:

With 3.0, the layout has changed and metadata flows have changed. The checkpoint loader is now the checkpoint selector, which should provide some convenience and added stability. The LoRA blocker no longer requires fiddling with the spaghetti string to toggle it, there is now a toggle switch. Sliders have been implemented throughout the Input center, including controls for the Upscaler.

New with 2.2:

Added toggles for including the workflow in the image's metadata.

Added a means by which you can bypass the upscaler and still send the image (directly from phase 1 gen/hires fix) to the detailer, rather than being forced to use upscaler in order to access the detailer.

Color Matching directly following Upscaler - I have found that some models have noticeable drift in color tone and this corrects that, works beautifully.

New with 2.1:

Detailer moved to the back of the order, after upscale, to ensure that full control over expressions remains with you, the artist. I'm very happy with this workflow in its current form, although there might be new versions or changes down the line.

New with 2.0:

Raise CFG without crushing the image

Upscaler with a third save node

Somehow resists LoRA bleed, whether it's the scheduler / sampler combo or what I'd love to know. It's not perfect though!

Preview KSampler - a game changer for skipping unsatisfactory seeds. Use it!

You will need to download an upscaler model if you don't have one already.

How this Workflow Came to Be

I was using one of these massive workflows with all the bells and whistles and I ran into a snag. A LoRA was confounding the hand detailer and the hands were all coming out like macaroni alfredo. So, I turned off that LoRA. I was also running into a common problem: when two characters are in the frame, they end up with the same eye or hair color. The tags were blending across character BREAKs which is ultra common, and detailers that handle all bboxs in order cannot be instructed on each bbox individually. That's something that my I2I workflow was designed to do, and the mega huge T2I workflow with all its heft didn't have detailer separation as a feature.

I don't think I could add to such a complex workflow without tripping up and it end up taking 2 hours. I thought if I'm going to modify a workflow it might as well be my own. It would keep things simple with only nodes I know I have, I'd finish in under 30 minutes, and then I would have complete control over faces.

Saving the metadata is important too, however somehow in converting things and removing things that get concatenated (such as the natural language prompt), the metadata portion got a little scrambulated, and don't worry I fixed it -- but 30 minutes turned into a several hour affair. Under nine hours though, I'm pretty sure. The latter part was mostly testing and tweaking. I ended up making some changes to the detailer modules so that they don't spam the image feed, and upgraded that module to V1.1.

During testing, I started with one figure, then two, and I soon discovered that this workflow is absolutely a beast at ensuring your faces all look great. Whether you have one face in the image or five, the detailer can make each one pop with high aesthetic quality and unique expressions, and the SEGS previewer keeps you in control.

The hand detailer gives you per-segs control over hands although sometimes bad hands are going to happen, particularly when there are many in one image. Fingers in AI is always a little trickier, but the workflow allows you to add a hand detailer LoRA directly to the detailer, upping your chances of success dramatically. Do it. At the same time, that LoRA will be blocking other LoRAs that could create havoc on delicate fingers. I haven't seen this approach elsewhere, curious whether it holds up in your experience. Please let me know in the comments what you think about my LoRA blocker, if you have any aspects you think ought to be changed, bug fixes, etc. This is just version 1, it's bound to evolve.

Requirements

You will need a face bbox detector - like this one

You will needs a hands bbox detector - like this one

You might decide to add or replace a bbox detector for face or hands with one for eyes, or add another bbox array. For eyes, try this one

In ComfyUI you may need to turn some safety features off to use pickletensors. Consult AI if you run into problems.

Custom Nodes Used in the current version:

ComfyUI Impact Pack

ComfyUI_lc123_nodes

rgthree-comfy

ComfyUI-KJNodes

ComfyUI-Easy-Use

ComfyUI-Crystools

efficiency-nodes-comfyui

Comfyroll Studio

Dynamic Thresholding (SDXL Only)

ComfyUI-Lora-Manager

ComfyUI_essentials

CG Use Everywhere Anything Everywhere

ComfyUI Impact Subpack

ComfyUI-mxToolkit

Skimmed_CFG (SDXL Only)

Comfyui_Mira

ComfyMath

ComfyUI Image Saver

ComfyUI-LogicUtils

JPS Custom Nodes for ComfyUI (Anima/Krea)

ComfyUI-Krea2T-Enhancer (Krea Only)

Rebalance Pack (Krea Only)

Files with SHA256: ef482ab0d830e6a12422270a8a55e4e19ba1e10744f15ff106de37b3f13a26a2