Update Highlights – Version 1.1
Improved overall readability and interface for a cleaner, more user-friendly experience
Completely reworked the Highres Fix section — now toggleable for greater control and flexibility
Added automatic metadata embedding in the final generated image
Key Features:
▫ Multi-Image Story Framework
Generate up to 8 images at once, split into 3 sections for natural storytelling flow
Ensures visual consistency across panels for a true comic/manga feel
User-friendly separation of prompts → build your character from A to Z with clarity
▫ Wildcards & LoRA Integration
Full support for wildcards to add variation and replayability
Smart LoRA support for style, character, outfits, and accessories
Flexible structure to mix & match without breaking consistency
▫ Built-In Comic Page Assembler
Automatically aligns generated images into a single comic-style page
Perfect for previews, storytelling drafts, or final page layouts
▫ High-Quality Image Tools
Highres Fix script for crisp, detailed outputs
Optimized node flow keeps generations stable while maintaining detail
▫ Intuitive Prompt Design
Prompts are cleanly separated for background, character, details
Easy to understand and tweak, even for beginners
Designed to stay user-friendly while allowing advanced customization
▫ Perfect For:
Storytellers building short comics or illustrated sequences
Artists wanting consistent characters across panels
Creators exploring manga-style or visual novel outputs
▫ List of custom nodes used :
How to use :
https://civarchive.com/articles/20162
Join the Discord!
Come hang out, share your AI art, give feedback, ask questions, or just vibe with the community.
It’s also the best place for troubleshooting, commissions, and sneak peeks of upcoming work.
Description
FAQ
Comments (32)
Olá saudações do Brasil... seu workflow é maravilhoso... mas o highresfix script apresenta erro >> Prompt outputs failed validation: Value not in list: modelname: 'waiNSFWIllustrious_v120.safetensors' not in (list of length 69) HighRes-Fix Script: - Value not in list: control_net_name: 'control_v11p_sd15_openpose.pth' not in []
sou novato nao sei muito bem o que tem que fazer perdao se for muito facil...
Hello, greetings from Brazil! Thanks a lot, I’m really glad you like the workflow 🙏
The error you’re seeing isn’t actually a problem with the workflow itself — it just means ComfyUI can’t find the exact model/checkpoint or ControlNet you selected:
Model error → waiNSFWIllustrious_v120.safetensors isn’t in your models folder (or it’s named differently).
ControlNet error → control_v11p_sd15_openpose.pth is missing, so ComfyUI can’t load it.
To fix it:
Make sure the model file you want is downloaded and placed inside your ComfyUI/models/checkpoints/ folder, with the exact same filename as in the workflow.
For ControlNet, download the right .pth (e.g. control_v11p_sd15_openpose.pth) and put it inside ComfyUI/models/controlnet/.
Restart ComfyUI so it detects them.
If you don’t need those specific ones, you can just replace them in the workflow with a model and ControlNet you already have.
Don’t worry about being new — these kinds of errors happen to everyone at the start!
The workflow generates many errors that I can't solve.
Well ? If want me to help your going to need to give the errore code !
I'm getting this error: "When loading the graph, the following node types were not found
SDAnyConverter" , i search in custom nodes menu on comfyui and try to fix, but it never works.
Go to the GitHub repo of the custom node (for example: https://github.com/username/ComfyUI-SD-AnyNode).
Click the green Code button → Download ZIP.
Extract the ZIP.
Place the extracted folder inside your ComfyUI/custom_nodes/ directory.
👉 Full path should look like:
ComfyUI/custom_nodes/ComfyUI-SD-AnyNode/
Restart ComfyUI completely.
After restart, ComfyUI should load the new nodes and the error about SDAnyConverter will be gone.
@Mittoshura Thanks, i'll try that tomorrow, really intrigued with this workflow.
@Mittoshura Not working, i tried to clone any node that could possibly be related to the error, still shows up, so this is unusable. Any other node that i could use in exchange of this?: https://ibb.co/Hfvn4rN8
@skyrimer3d You can just delete it and put the same scheduler manually on all sampler ! It's not the most convinient but it will work I think !
I personnaly use sgm_uniform !
@Mittoshura i'll try that thanks
@Mittoshura No way, i tried changing it to a scheduler selector node, tried both "normal" and "sgm_uniform", but then i'm getting tons of errors : (Prompt outputs failed validation: Return type mismatch between linked nodes: scheduler, received_type(['simple', 'sgm_uniform', 'karras', 'exponential', 'ddim_uniform', 'beta', 'normal', 'linear_quadratic', 'kl_optimal', 'bong_tangent']) ... .) .
@skyrimer3d Did you try to input it manually in the sampler node, don't create another node?
@Mittoshura I managed to fix it by switching the node to "any type switcher" instead, cnr id "comfyui_llm_party". Now i'm getting this: https://ibb.co/6cQtKppq . I've disabled the high-res fix script node and it seems to start working, it's already producing images (i didn't change anything so a pic of a guy with his d*ck off showed up lol), so it seems to be working, hurray!
@skyrimer3d Great progress! The issue with the HighRes node might be that it's looking for a specific model you don't have installed. To try and fix it, right-click on the node and select both 'Fix node v2' and 'Reload node'. Also, try setting the seed value to -1. This should force it to use a random seed and might resolve the conflict.
@Mittoshura I tried that but still the same error. The images look good enough, so i'm not missing the upscaling much.
Hi @Mittoshura this workflow looks amazing but I started with it today and cant get past this error. KSampler (Efficient) 'UnetModel object has not attribute x_embedder chat gpt says its because i have a mix of Sd3.5 and Sdxl Unets, Im relatively new to all this, I guess I should start by asking if this workflow can run sdxl models?
Thank you!
Hey! This workflow isn’t really made for SDXL — it’s designed to work best with Illustrious or Pony models.
That’s why you’re getting the Unet mismatch error.
Try switching to one of those and it should run fine!
Hey this is awesome! any chance you have a mini guide that explains how to use the prompts inputs effectively. Trying to understand how to use the 3 additional prompt boxes with headings like "preliminaries"
Each section (SFW, preliminaries, and intercourse) has its own global prompt that will flow into every image within that section. For example, if you want all three images in the preliminaries set to include “baseball bat,” just add it to the global prompt of that section.
Then, you can refine each individual image prompt to add variation and extra detail. This approach is especially powerful when combined with wildcards!
@Mittoshura Thank you, i've managed to run your workflow well and pretty impressed with the results. just a feedback and i dont know if this is possible but can we set separate resolutions for each image generated. Some resolutions can give better results for certain view styles.
@JSaitoshin Yes that should be really easy in fact ! You just have to input a different latent node in each sampler ! Do you know how to do it?
@Mittoshura im not sure how to tbh, im still learning how to use comfyui and create workflows. some guidance would be appreciated
@Mittoshura i figured it out. its just the saved images gets scaled to a fixed resolution. im gonna have to figure that out
You owuldnt be kind enough to share maybe your wildcards right? It would make it a lot easier to see how you structed them and take some inspiration!
Of course ! It's not ordered but do want me to make a zip and send it to you ?
@Mittoshura That would be fantastic!
Here you go ! I've uploaded it here : https://civitai.com/models/1937755?modelVersionId=2193151
@Mittoshura you, very kind and helpful!
Sd any converter failed import (( sad
The problem comes from the any converter that input the scheduler name in the sampler, try to delete and input it manually and it should work !