Learning AI image generation implies a simultaneous discovering of technology, tools and methodologies inside a field which is evolving with unprecedented speed.
Here is a collection of different explorations.
Here is a collection of different explorations.
CARAVAGGIO MEETS OTOMO
This experiment's focus it about understanding how the model (Deliberate 1.0 a dream-booth of Stable Diffusion 1.5) mixes and matches two different artist's styles and motives.
All prompts are based on the following positive conditions: concept art (rather than picture or painting), female body (in order to avoid portraits and/close up), technology (to incorporate sci-fi elements); whereas the negative ones are about: environment (to better focus on the character), and all the usual tips to avoid deformed limbs or faces.
All prompts are based on the following positive conditions: concept art (rather than picture or painting), female body (in order to avoid portraits and/close up), technology (to incorporate sci-fi elements); whereas the negative ones are about: environment (to better focus on the character), and all the usual tips to avoid deformed limbs or faces.
DIFFERENCES IN ARTISTS' INFLUENCE
Given the same seed and the same prompt, the different artists' weights influence the outcome, by incorporating and/or discarding specific visual features.
Otomo ++: oriental facial features, straight hair, more angry and/or smiling characters, more technological components, slimmer body; flatter image, contours, less shades (sometimes cell shading).
Caravaggio ++: western facial feature, curly hairs, more serious and/or melancholic characters, minor technological prominence, chubbier body; nearly no outlines, more "chiaroscuro", blurred areas (sometimes brush strokes).
Differently from most recent models (ST XL, 2023), this one cannot draw a neat distinction between the subject and how it it is represented (visual style). This means that the more prominent the artist parameter is (in a 0 to 1.5 scale), the more the artist's specific iconography, as well as his stereotypes, are likely to take place.
Differently from most recent models (ST XL, 2023), this one cannot draw a neat distinction between the subject and how it it is represented (visual style). This means that the more prominent the artist parameter is (in a 0 to 1.5 scale), the more the artist's specific iconography, as well as his stereotypes, are likely to take place.
When Caravaggio parameters are the highest ones.
When Katsuiro Otomo's parameter prevails.