Konstantin Korovin. Training a Generative Neural Network in the Style of an Artist

Konstantin Korovin—one of the leading Russian Impressionists, a «master of light and the moment.» He painted with a rapid, vibrating brushstroke, favored rich, almost neon color, backlighting, and reflections on water and glass, and worked within a very recognizable range of subjects: night Paris, the Crimean resort of Gurzuf and its terrace overlooking the sea, still lifes of fishermen, manor verandas, portraits of women in the garden, and theatrical sets.

The series of 12 images is structured as a «year-in-review» diary: each «canvas» represents a month, and along with the month, the location/narrative shifts, mirroring Korovin’s real creative trajectory. The viewer cycles through the artist’s geography (Russia $\to$ Crimea $\to$ Paris $\to$ back to Russia), while the changing seasons provide a natural drama of light—from a frosty winter morning to a warm southern night and back to snow. This allows for a demonstration of the trained model’s capabilities across the full spectrum: cool and warm light, interior and plein air, still life, landscape, portrait, and theatrical set design—thus proving that the style remains consistent across varied subject matter, rather than being confined to a single type of scene.