Original size 1140x1600

Learning Stable Diffusion in the style of Leonardo da Vinci

Longread translated automatically
The project is taking part in the competition
big
Original size 2000x700

Project Idea

Blah blah blah

big
Original size 2110x422

For model training, I compiled a dataset of 20 hand-painted oil paintings by Leonardo da Vinci. By choosing Da Vinci specifically, I ensure that his works are not subject to copyright and are considered public domain.

Model Training

The model training process was performed on Kaggle, utilizing Fine-tuning Stable Diffusion XL and the LoRA method. For the model’s training itself, 20 square images, each at a resolution of 1280×1280, were used.

During the model fine-tuning, I encountered several issues not mentioned in the course material. Because of this, I needed to add several extra lines of code and work with two T4 GPU accelerators, rather than the P100 GPUs specified in the lectures.

Result, Series

0

— oil painting in DAVINCI style, angelic woman with long hair
–oil painting in DAVINCI style, angelic man with long hair
–oil painting in DAVINCI style, young angel

Conclusions

Code + Source Files

Original size 2000x700
Learning Stable Diffusion in the style of Leonardo da Vinci
Project created at 26.09.2026