Clip specific materials.
■Material test images of spheres covered with fabric-like layers, like the sample, are quite common.
Sometimes you just can't help but wonder what's underneath. So I thought using a LoRA to remove the covering would resolve that curiosity. Now, there's no need to keep wondering about it.
■This LoRA was created for editing anime-style images.
■Details and prompts can be found in the shared workflow.
■If you're new to Kontext, try installing it using the instructions at the URL below.
It's well-documented and easy to set up.
https://docs.comfy.org/tutorials/flux/flux-1-kontext-dev
■This was trained 2D illustration images.
The scale is small, but the detection rate is surprisingly high.
■If you're not satisfied with the texture or detail after clipping, try i2i or inpainting with another model.
SD1.5 is lightweight and works well as a high-resolution refiner—recommended for i2i. It only uses about 2GB of VRAM, but it's very powerful.
It performs reliably up to 1024×1536px.
Try denoise values between 0.25–0.5
■If you're unsure which SD1.5 model to use, feel free to try my merged SD1.5 models first.
I provide both real and anime styles—use whichever fits your image better.
If the results don’t match your expectations, you can then look for a model closer to your ideal!
https://civarchive.com/models/1246353/sd15modellab
■If you have enough VRAM, SDXL could also be a good option.
■I used AI Toolkit to train this LoRA.
If you're interested in training, the developer has provided a tutorial at the URL below — give it a try!
I think you'll find it's easier than you might expect.
Description
トレーニングデータ( 8.84 KB ):comfyui workflow
FAQ
Comments (17)
Was the action to remove the material on the top side?
It's a undress model for anime pictures and the triggerword is "remove"?
Well, just to be safe, I won't say anything too direct—but if you look at the workflow, you can probably guess what's going on lol.
@hjhf if u add that on tensor art for a Dollar people would buy it because its hard to find a Lora for Flux Kontext these days.I got so many ideas but I'm just to lazy such as like nipple pokes.Areola slips.Nipple slips.Remove bra outline
@Santaonholidays
Thanks for the info!
I haven't used Tensor Art, but I’ll consider uploading there in the future.
LoRA training with Kontext is super easy—I hope more people give it a try.
It's much simpler than most think: just train on visual differences.
No need to overthink—it's predictable, easy, and results are clear.
Even with just a few dozen images, you’ll notice effects in about 1000 steps.
And with a day of training, you can get surprisingly usable results.
@hjhf What's the captioning and dataset look like for that? Are the images submitted side-by-side with direct reference to the image? like image 1 description & image 2 description, or cause-and-effect? like image 1 description & image 2 differences ('now' words, etc).
@zefy
There are two datasets: one with the original images and one with the edited results.
You only need to add the instruction prompt (what you want to achieve) as a caption to the edited images.
Note: This method works with AI Toolkit, but may differ in other tools.
I think the developer's tutorial makes the process easy to understand.
https://www.youtube.com/watch?v=WSWubJ4eFqI
You might not even need my explanation after watching the video, but just in case, I’ve outlined my workflow below for reference.
Example:
Instruction: "Remove the person from the image."
Original image(control side): with a person
Edited image(target side): without the person
(You can either remove the person or add one to a blank image—whichever is easier.)
Only the edited image needs a caption. The original doesn’t.
There may also be cases where the target you want to remove isn’t limited to just people or a single object.
While specific prompts may help, broad ones often work fine too.
Example: "Remove people, cars, animals, and any objects disrupting the cityscape."
If you want more accurate captions, tools like JoyCaption can help—just ask it to describe the image and end with something like "Please remove the person from this image."
@hjhf Hmm, how would it know which image to pair it with? I would imagine maybe a minimal caption for the source set might make sense, even just for an identifier to pair the changed image with. Thanks, now I just need to motivate myself to train something. I lack motivation to an incredible amount!
@zefy
I'm not sure if this fully answers your question, but I believe pairing is based on matching filenames. You need two folders—one for original images and one for edited images—with matching filenames. For example:
Originals > 001.jpg
Edits > 001.jpg + 001.txt (instruction prompt)
This lets the system learn the difference. In AI Toolkit, the .txt is auto-created via the GUI when adding captions, so we don’t need to worry much about it.
It’s definitely hard to decide what to make. Something like the following should be easy to create.
when you already have before/after image pairs—like facial expressions, day/night backgrounds, or seasonal changes.
You can also prepare materials in reverse, like applying blur to create training data for a deblurring LoRA.
Once you try one, you might get ideas for more. Good luck!
Thank you! I'll give it a try as well.
what material clipper means !?!
There are many images of cloth-covered spheres like the sample. Sometimes you just can't help but wonder what's underneath. I made a LoRA to cut it open and solve that mystery. lol
Maybe I should’ve just written that in the description.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.