Vision Vanguard 22 [Krea 2]
Cinematic high-contrast realism with deep low-key shadows, detailed textures and precise depth of field.
Base model: Krea 2
Trigger words: none required
Helpful keywords: cinematic photography, high contrast, shallow depth of field, deep shadows, natural lighting
Strength: 0.85–1.0 (1.0 for maximum grit, 0.85 for smoother skin)
Negative prompt (optional): flat lighting, low resolution, plastic skin, cartoon, anime
Prompting: subject (age, origin, expression) → action → environment and materials → lighting direction → mood.
Tips
Texture words like textured, weathered, rippled, gnarled and wet bring out surface detail.
It handles low-key scenes well. Black backgrounds and deep shadows are fine.
Use low angles for drama and high overhead angles for abstract compositions.
Combine muted tones with a single vivid accent.
Description
Base model: Krea 2
Strength: 0.85–1.0
Resolution: trained at 512, generate at 1024
CFG: 2.5–4.5 (Base) / 1.0 (Turbo)
Steps: 35–50 (Base) / 8–9 (Turbo)
FAQ
Comments (2)
Each one is amazing, my only complaint is that I wish they had different easier to remember names, so I could know what they do differently, at least without having to refer to the pages.
Hey. Thanks. Yeah I'll make different LoRAs which actually do different things and those will be named differently. I'm just shortly working through a bit of a backlog training wise. And I'm still working on that 80s 90s etc decade dataset.