Traditional Korean Painting Model

(Translated from English using DeepL)

I renamed the model because I felt it was creating unnecessary misunderstandings. I fine-tuned it with data from Joseon era painters, modern Korean paintings, and ink paintings.

Hypernetworks and SD1.5-based model were created a few months prior. Then, SDXL appeared. At that time, I did not work on the SD2.1-based model because I was hesitant about using ControlNet, and because it wasn’t clearly superior to SD1.5. While prompt-based changes are clearly achievable, and reproduction of known subjects looks slightly improved, there is no clear advantage over the higher specifications. However, with SDXL, ControlNet and various other features were said to be supported out of the box, and the hardware requirements were not harsh for the supported resolutions, so I decided to give it a try.

For more information on how it was created, see this article.

HuggingFace Link

https://huggingface.co/gagong/Traditional-Korean-Painting-Model-v2.0

https://huggingface.co/gagong/korean-sumukhwa-model-ver-1

Used data

Joseon Dynasty Painters Link

Because the number of paintings by Joseon-era artists is limited, I increased the dataset size by fragmenting the images and applying horizontal flips. Total of 5,684 images.

Digital K-art data Link

I reduced the dataset from 11,246 to 6,253 images. The AIHUB data is immense, so I first filtered it down to labeled data, resulting in 11,246 images, then selected those that were close to square-shaped, leaving me with 6,253.

Production data for traditional Korean ink paintings by style Link

I used 5,860 images cropped into squares. Though this dataset is very large, I kept the quantity similar to that of the Joseon-era painter data.

A sample of this data is shown below.

User Guide (SDXL version)

Usage Tips

1. In general, higher CFG scale settings lead to worse quality. We do not recommend setting it above 10. However, shorter words seem largely unaffected.

2. The longer the input sentence, the more likely the output will resemble a generic photo. Increasing the weight of trigger words does not improve the result.

3. The recommended CFG scale is between 4 and 8.

4. Do not use low-quality trigger words in isolation.

5. While I haven't tested using multiple trigger words simultaneously, it's likely to yield better results in general.

6. If the output appears too photo-like, try combining the prompts "painting" or "style." For example: kimhongdo painting

7. Since this hasn't been used extensively yet, the recommended methods may be revised in the future.

Trigger Words List

Note that this does not accurately reflect the true painting style of the artists.

View as a larger image Link

• whtjs
  Emulates the average painting style of a Joseon era painter.

• kangsehwang
  Drawn in its simplest form.

• kimhongdo
  Highest quality.

• sinyunbok
  Second highest quality.

• simsajeong
  Higher probability of a woman’s face appearing in the image.

• anjungsik
  Lowest quality.

• jangseungeop
  Second worst quality.

• heoryeon
  Produces results with the least color.

• gksrnr
  Creates modern forms. Body errors are frequent.

• tnanr
  Produces an average visual appearance of the three ink painting techniques below.

• baengmyo
  Often yields results similar to those of ‘whtjs’.

• gureuk
  Results in images with a thin border appearance.

• molgol
  Produces results with no visible boundaries.

User Guide (SD 1.5 version)

Trigger Words: GKSRNRGHK, RLAGHDEH

Approximately 6,000 images of Korean ink wash paintings and around 1,000 images of Kim Hong-do’s works were used. Because of this, for the Korean ink wash style, CFG Scale between 2–7 is recommended, and for Kim Hong-do’s style, between 4–12. If using both styles, it's best to use a mid-range value. The step count also influences the result, so choosing an appropriate value is essential.

Although performance is limited, technique-specific prompts are applied, and you can use the following prompts when using the model. (However, the differences observed are subtle.)

• Baekmukbeob (White ink technique): baegmyobeob

• Molgolbeob (No-boundary technique): molgolbeob

• Guleugbeob (Thin border technique): guleugbeob

To emphasize Kim Hong-do's style, use both “rlaghdeh style” and “rlaghdeh painting” prompts.

Sample Images

This is a sample image from txt2img.