Styles
Last updated
Last updated
FLORA’s Styles feature allows you to apply specific visual styles to your image generations. Styles Feature enhances t2i (text-to-image) and i2i (image-to-image) generation by allowing users to apply predefined aesthetic filters or train their own custom styles. This feature enables greater creative control, ensuring that outputs align with specific artistic or brand identities.
FLORA provides a curated selection of predefined styles, optimized for high-quality and consistent results across different image generations. These styles modify the aesthetics, color, mood, tone, and content (characters, objects) of generated images.
Users can choose from a growing library of styles, categorized as follows:
Fisheye
Depth cues, model shading, digital flesh. Not about real-world cameras—more like Blender renders. LoRA captures things like subsurface scattering, ambient occlusion, baked shadows. Geometry feels synthetic, form takes priority over texture.
3D
Depth cues, model shading, digital flesh. Not about real-world cameras—more like Blender renders. LoRA captures things like subsurface scattering, ambient occlusion, baked shadows. Geometry feels synthetic, form takes priority over texture.
X-Ray
Imagined transparency. Subjects are peeled open—bones, wiring, inner systems exposed. A LoRA that trains on radiographic motifs but fakes the effect across arbitrary subjects. Useful for metaphor: revealing the hidden.
Studio Lighting
Controlled shadows, sharp highlights. Key light, fill light, rim light—classic three-point setup. Every pixel feels considered. LoRA adapts subject matter to sit inside a constructed scene, often with high contrast and shallow depth of field.
Motion Blur
Speed encoded into the image. Simulates long exposure or rapid motion. Directional streaking, trailing forms. Often used in cyberpunk, dance, or action-heavy outputs. The blur is compositional—it pulls the eye.
Monochrome
One hue, infinite values. LoRA removes color information, but enhances detail through contrast. Could be charcoal, cyanotype, or film noir. Emphasizes shape, light, emotion—less distraction, more weight.
Architectural
Structured, rectilinear, scale-aware. Linearity dominates—LoRA learns blueprints, elevations, orthographic projection. Often defaults to modernism unless nudged. Useful for clean geometry and visual order.
Soft Focus
Dream logic. Edges melt. Light blooms. Often draws from portraiture, especially analog. Feels cinematic or romantic. This LoRA suppresses crisp detail in favor of mood.
Animated
Stylized, minimal texture, exaggerated form. Pulls from anime, Western animation, or game cutscenes. Flat shading, color blocking, large eyes, expressive lines. A LoRA that suppresses photorealism in favor of symbol.
Vector
Infinitely scalable. Line clarity, color fields, no noise. Think Figma assets, logo packs, poster icons. A LoRA that favors SVG-like outcomes. It flattens depth, prioritizes readability.
Photorealistic
As close to lens-based capture as possible. High fidelity textures, correct shadows, plausible lighting, skin imperfections. LoRA draws on real-world datasets and DSLR optics. Often pairs well with materiality prompts.
Line Drawing
Gesture over mass. Black ink, pencil, or digital stroke. Negative space plays a role. LoRA reduces values to contours—ideal for diagrams, instructions, or ideation sketches.
Product Mockup
Perfect surfaces. Rendered packaging, floating objects, top-down minimal backdrops. LoRA learns from ecommerce, product design, and pitch decks. Prioritizes clean, centered, desirable.
Extreme Detailer
Texture fanatics. Zoom in: every pore, crack, wrinkle. This LoRA overfits on microstructures—hyperreal to the point of obsession. Can verge on the grotesque if unchecked.
Cinematic
Frame as story. Aspect ratio matters. Lighting dramatizes. Color grading carries tone. LoRA may reflect film stock, composition theory, or golden hour. Meant for stills that feel like scenes.
Each style adjusts image attributes such as aesthetics, composition, mood, tone, and content.
For users seeking high-level, customized creative ownership, FLORA enables custom style training, allowing users to upload reference images and train a model that adapts to their unique aesthetic.
Upload Reference Images – Provide a set of images that represent the desired visual style. Hit Create Style button to start the training.
Training Process – When the training is done, you'll get a notification.
Preview – You can find your customized Style on the left side tool bar. Your custom style will also show up on the tool bar above the image block, with all the other default styles.
