PanelPainter-Project is an open-source initiative to automate black-and-white manga coloring using fine-tuned LoRAs. This project is dedicated to training models that maintain clean line art while achieving smooth, anime-style color fills.
Update – V3.0 Major Release
PanelPainter V3 is a major milestone. Unlike previous "helper" versions, V3 is a standalone coloring lora trained on Qwen Image Edit 2511.
Training Data: Trained on 903 hand-picked panels using Natural Language Captions.
Methodology: Combines the "real line art" training method discovered in V2 with a significantly larger dataset to generalize across different manga styles.
Trigger Word:
Color this panelpainter
Workflows & Resources
For the best results, use the dedicated workflows on RunningHub.
PanelPainter V3 (Qwen 2511)
RunningHub: V3 BF16 Workflow (Fast, Balanced)
RunningHub: V3 AIO Workflow (All-In-One)
PanelPainter V2.5 (Qwen 2511)
RunningHub: V2.5 AIO App (With VL Prompting) – Includes Vision Language prompting for better adherence.
Usage Guide
V3 has been tested on various styles including Chainsaw Man, Frieren, Komi Can't Communicate, and Oshi no Ko.
Recommended Generation Settings
LoRA: PanelPainter V3 (Weight: 1.0)
Helper LoRA: 4-Step Lighting (Weight: 1.0)
Steps: 4
Sampler: Euler
Scheduler: Simple
CFG: 1.0
Prompting Use the trigger word in your prompt:
Color this panelpainter
You can also add specific lighting or atmospheric tags:
Color this panelpainter, sunset lighting, warm tones
Project History & Development
Version 3.0 (Current)
Base: Qwen 2511
Summary: Scaled up to 903 images using the "real line art" training method. Solves the generalization issues of V2 while maintaining color quality.
Version 2.0 (Stable)
Base: Qwen 2509 (Compatible with 2511)
Summary: The breakthrough version. Switched from synthetic data to real line art (150 images). Proved that small, high-quality datasets outperform massive synthetic ones.
Version 1.0 (Legacy)
Summary: Trained on 7,000 synthetic grayscale images. Failed to handle real ink imperfections. Kept for archival purposes.
Credits & Acknowledgements
Training: Trained on Musubi Tuner (Thanks to kohya-ss).
Dataset Contributors: Special thanks to @Rox_Jr & @lucifer_brine04 for their help with dataset curation.
Description
Version 3.0 Change Log
Standalone Release: Now a fully standalone coloring model fine-tuned on Qwen Image Edit 2511.
Dataset Upgrade: Trained on 903 hand-picked panels (50% Doujin / 50% SFW) using Natural Language Captions.
Improvements: Significantly better generalization across different manga styles and fixed pixel shift issues.
Workflows: Added new V3 specific workflows for BF16 and All-In-One (AIO).
FAQ
Comments (20)
Hi please help , NOTHING is happening with the coloring, the output image becomes all black apparently... i tried with strength 0.5 and strength 1 for your Lora.
Im using PanelPainter_v3_Qwen2511 for Lora,
and Qwen-Rapid-AIO-NSFW-v18 , for the checkpoint
I tried using your PanelPainter_V3_QIE_2511_AIO.json for the workflow (latest )
But yeah nothing works after i generate... the prompt used is unchanged : Color this panelpainter
Pls help, this doesnt make any sense why it doesnt color..
Try running ComfyUI with --precision-full to see if that helps. People with the same issue said it worked for them.
@Kokoboy Thx for the reply, but sadly still no luck... maybe i didnt write it correctly in the bat file? I have my comfyUI bat file set up like this :
.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --use-sage-attention --fast
set COMMANDLINE_ARGS=--precision-full
pause
..is this the correct way to add that precision full command line? sorry im not familiar with this more technical stuff. With the above commands it still shows a black bar when i try coloring a page.. :/
ok for anyone having this issue i FINALLY found the solution!
You have to REMOVE that : --use-sage-attention --fast
line if you have it on the bat file... now the colors work well!
I followed all the steps, including uploading the Loras file to the checkpoint and uploading my image for coloring, but when I clicked run, a "Reconnecting" message appeared and my local machine shut down. I restarted it, but the problem persisted.
Did something happen to the link to the AIO model?
This sounds like a nice initiative, does someone know if there is a similar concept but From A base colored lineart to a Colored one (shades and details)?
As far i saw, nope
Can this be used for reference-based coloring? If so, are there any specific prompts for that purpose?
It doesn’t work very well, but you can try using a character close-up image as image reference with this prompt. It works to some extent, though the consistency is not great.
Prompt: “Color this panelpainter, refer to the character colors from the reference image.”
This model is extremely powerful and very user-friendly! However, Qwen Edit 2511 runs very slowly. Are there any plans to consider releasing a coloring LoRA for the Flux Klein version?
I do have a plan, but I’m saving my LoRA training budget for Z Image Omni / Edit Model LoRA training, which is why I’m not investing in Flux2 Klein at the moment. Although it’s a great model, I have high expectations for Z Image and believe it will significantly boost PanelPainter. Instead of training PanelPainter LoRA on Klein right now, I’m focusing on research and experimentation with the Klein model to improve PanelPainter’s reference consistency when i train a lora for it.
@Kokoboy Looking forward to your z-image lora
is there a way to generate faster.mine takes around to 60 seconds for Q4
Sadly nope
I used this Lora to create a batch coloring workflow, coloring all the comics in the entire folder, and shared the techniques for batch coloring multiple folders. Please refer to the article at the following link: https://civitai.com/articles/26062
This LoRA is amazing! The only issue is that the hair and skin colors change slightly from page to page, even when I specify them in the prompt. What would be the most logical solution for this?
Nice LORA! But will you release an illustrious version? I really need that!