🌀 ComfyUI Workflow: Character Pose & Expression Generator for LoRA Training
Ever wanted to train a LoRA but only had one good image of your character?
That’s exactly what this workflow is for.
It takes a single reference image and generates your character in tons of new poses, angles, and expressions — all while keeping them on-model. Perfect for quickly building a clean, varied dataset.
What it does:
✅ Creates dynamic body poses (standing, sitting, kneeling, jumping, flying, punching, kicking, etc.)
✅ Adds anime-style expressions (happy, sad, angry, smug, love-face with heart pupils, confused, yelling, crying, etc.)
✅ Includes hand + finger gestures (peace sign, pointing, fists, clasped hands, heart-shape, etc.)
✅ Keeps the style consistent with flat 2D tones + crisp line art
✅ Gives you ready-to-train outputs with minimal cleanup needed
Why use it?
LoRAs really shine when the dataset has variety — but collecting or editing images of your character is a pain. This workflow basically does the heavy lifting for you: drop in your character image, hit generate, and you’ll get a nice spread of poses and moods to train on.
I made it to help myself, but I figured it might help others who are building LoRAs too. If you try it out, let me know how it works for you — or share your own tweaks back!
Description
This version focus on 2D chars but can be easily tweaked to realistic
FAQ
Comments (18)
I tried using this to create a dataset for my upcoming character lora, it works but the face is almost always inaccurate. Is there a way around that or is that something im just gonna have to fix on my own with inpainting?
I completed dozens without any issues. Would you mind sending your character so I can try? If possible, one of your generated face with errors and what you think is wrong.
@alcaitiff of course, the first image is input photo, but theres times where a couple of outputs do not give the characters the black eye sclera's and black lipstick as shown in my input photo
I got it working! Your character is so cool that I'm generating all images and will post an album for you. The issue seems to be that Qwen doesn't know how to create nice black scleras. I added it to the positive prompt on the upscaler, as the other model handles it better. I also increased the denoising slightly. The best solution would be to train a black sclera LoRA for Qwen. WF -> https://pastebin.com/HUeVKHtp PICS-> https://imgur.com/a/QLN7GUf
@alcaitiff Thank you so much man, youre really a life saver!
@GGNerine you are welcome
If you feel comfortable, please post some of your results so others can better understand what the WF does and how it can help them.
@alcaitiff Will do man!
More images of your char, without the costume now -> https://imgur.com/a/zsdiItl
I will say as a sort of update, i get MUCH BETTER training data using the V1 prompt box than the other two.
great tool for Lora. but my chara is much too complex :p
i wanted to do that for a while but QWEN as powerful as it is doesn't give enough accuracy on fine details. thanks for going through the hurdle of making the workflow still :D.
some suggestions :
add top down view and from bellow aswell as back view.
i would also add an optionnal detailler pass after the upscaler but with IPAdapeter reference + lineart controlnet for the face.
--unless i find a way to inpaint the face with QWEN_EDIT acurately--
the same workflow for face/portrait only would be a much beter suit for me in completing the dataset.
Hi, thank you for the detailed comment and for your kind words. On this workflow you still can add the details on the prompts to guarantee your specific details. But the IPAdapter is a great idea. I'll create a version with it and post it here.
@alcaitiff oh since you're here ! what was that choice of sampler/scheduler (res_2s and bong_tangent) ? my ComfyUI don't have them. does it matter and how did you manage to add them ?
@Le_Fourbe They come with RES4LYF custom nodes, and in my experience, they significantly improve the results. However, they also take almost twice as long, so there's that.
@alcaitiff THANKS CHIEF :D !
Wonderful. Works incredibly well. Also very useful for creating expression packs for use in SillyTavern. Thank you for sharing this great workflow with us. Cheers.
Партия гордится тобой.
Кошку-жену уже выслали.