Abstract
This paper demonstrates the fusion of Kandinsky’s and Wu Guanzhong’s styles using AI, yielding Eastern and Western point-line-plane aesthetics. The authors present a method for blending diverse styles using Stable Diffusion and LoRA and subsequently apply the synthesized hybrid aesthetics in design. The optimized LoRA training workflow enhances the quality and controllability of generated images. The authors enable efficient generation of high-quality artistic images and concept designs through seamless interpolation of trained LoRA model of artists’ styles. This approach exhibits considerable controllability and scalability, serving as a valuable aid in the creative endeavors of artists and designers.
| Original language | English |
|---|---|
| Journal | Leonardo |
| DOIs | |
| Publication status | Published - 1 Apr 2026 |
| Externally published | Yes |
-
Steering Large Text-to-Image Model for Kandinsky Synthesis Through Preference-Based Prompt Optimization
Zhou, A.-L., Wu, W., Wang, Y.-A. & Zhang, K., 20 Apr 2025, Artificial Intelligence in Music, Sound, Art and Design (EvoMUSART 2025): International Conference on Computational Intelligence in Music, Sound, Art and Design (Part of EvoStar) . Springer, p. 417-433 17 p.Research output: Chapter in Book or Report/Conference proceeding › Chapter › peer-review
Open Access1 Citation (Scopus) -
Human Aesthetic Preference-Based Large Text-to-Image Model Personalization: Kandinsky Generation as an Example
Zhou, A.-L., Wang, Y.-A., Wu, W. & Zhang, K., 9 Feb 2024, (E-pub ahead of print) p. 1-9. 9 p.Research output: Contribution to conference › Paper
Open Access -
Kandinsky as You Preferred
Zhou, A.-L., Wang, Y.-A., Wu, W. & Zhang, K., 25 Jul 2024, SIGGRAPH '24: ACM SIGGRAPH 2024 Posters. Association for Computing Machinery (ACM), p. 1-2 2 p. 15Research output: Chapter in Book or Report/Conference proceeding › Conference Proceeding › peer-review
Open Access2 Citations (Scopus)
Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver