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
8/21: I made some minor improvements to the workflow prompt.
I’ve addressed the main issues from previous versions.
This version produces results somewhere between v1 and v2.
Version Comparison:
v1.0 : Aesthetics: 5 / Detection Accuracy: 2
v2.0 : Aesthetics: 2 / Detection Accuracy: 5
v3.0 : Aesthetics: 4 / Detection Accuracy: 4 or 5?
v1 had great texture but poor detection.
v2 improved detection with more data but lost style quality.
v3 uses a refined dataset and larger batch size, improving both quality and accuracy. It’s also sharper than previous versions.
I recommend using either v3 or v1, depending on your preference.
v3 is a safer choice, but v1 can still produce more striking results.
Alternatively, try merging the LoRAs at 0.5 each — I’ve included that workflow too.
FAQ
Comments (4)
the far how this can be good is how it can predict the original material and shape how it looks in real as possible , not all woods are the same , as it can be closeley determine how the object is look if removed the thing from on it, the more real data to feed it for train will be better.
I tried using it on wan 2.1. Nothing seems to happen. and yes the lora file is added. But do you have a certain prompt you type in?
For the prompts, please check the Training Data files under the LoRA—those include the exact prompts used in the workflow. You can download them here:
https://civitai.com/api/download/models/2087406?type=Training%20Data
I realize this may feel a bit roundabout, but it’s an intentional precaution—thanks for understanding.
This LoRA itself isn’t meant to be used directly with wan 2.1, but if you run it with kontext-dev, it should work even on images produced by wan 2.1. You can find installation steps here:
lol wan.. bro are you like mentally handicapped or just retarded
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.