date
type
status
slug
summary
tags
category
icon
password
URL
Last August, I compiled a dataset of 6,500 images using public domain traditional Chinese paintings from various dynasties and trained a Chinese painting model using StyleGAN XL. Later, with the advent of large-scale image generation models like DALL·E 2, MidJourney (MJ), and Stable Diffusion (SD), the StyleGAN series, which had poorer generalization and was more difficult to train, gradually faded from memory, including mine. Recently, I remembered that I still had this "antique" model stored on my hard drive. While Disco Diffusion may be outdated, it has become a well-known retro style; in contrast, StyleGAN XL barely had time to shine before becoming obsolete. Reflecting on this, I felt a bit nostalgic, so I decided to run the model again to generate an AI Chinese-style animation. Paired with music, the result felt quite good, as if paying homage to two bygone eras. It made me suddenly miss the good old days.
- 作者:Simon Shengyu Meng
- 链接:https://simonsy.net/article/stylegan-cnink-en
- 声明:本文采用 CC BY-NC-SA 4.0 许可协议,转载请注明出处。
相关文章
Microcosmic Universe

Project Title: One and Three Objects, and An Attempt at Exhausting the Object

One and three objects, and an attempt at exhausting the object

DreamGaussian: The Stable Diffusion Moment of AIGC 3D Generation

How I Used AI to Create a Promotional Video for Xiaomi's Daniel Arsham Limited Edition Smartphone

Works Series - Dimensional Recasting
