CivArchive
    YFG Obs - Krea 2 - v1.0
    Preview 1
    Preview 2
    Preview 3
    Preview 4
    Preview 5
    Preview 6
    Preview 7
    Preview 8
    Preview 9
    Preview 10
    Preview 11
    Preview 12
    Preview 13
    Preview 14
    Preview 15
    Preview 16
    Preview 17
    Preview 18

    Description

    ( Not my Model )

    https://civitai.com/models/2413599/yfg-obs-flux?modelVersionId=3114935

    ๐Ÿ”ฅ Overview

    YFG Obs is a dark, cinematic, high-contrast mood LoRA built to inject controlled atmospheric intensity into your generations.

    This model pushes:

    • Deep, rich shadows

    • Bold contrast curves

    • Subject isolation / hero framing

    • Industrial / dystopian undertones

    • Noir + tactical composition bias

    • Cinematic lighting language

    Itโ€™s not just a โ€œcolor grade.โ€
    YFG Obs reshapes the lighting and mood structure of the scene while keeping the underlying subject coherent.

    The result is confident, intentionally dark , and stable โ€”even when you lean into it.


    ๐Ÿง  Model Behavior

    YFG Obs delivers:

    • Strong tonal depth

    • Controlled black levels

    • Clean subject separation

    • Atmospheric density

    • Cinematic shadow shaping

    Despite its strength, it remains surprisingly stable across a wide weight range.

    It typically does not :

    • Break anatomy

    • Collapse faces

    • Add random artifact noise

    • Destroy prompt structure

    Itโ€™s aggressive stylistically , but technically stable .


    ๐ŸŽฏ Recommended LoRA Strength

    Extensively tested from:

    0.5 โ†’ 2.0
    Sweet Spot: 1.00 โ€“ 1.25

    • 0.5 โ†’ subtle cinematic influence

    • 0.75 โ†’ adaptive dark grading / gentle mood

    • 1.0 โ†’ strong Obs identity

    • 1.25 โ†’ confident, dominant mood

    • 1.5โ€“2.0 โ†’ heavy stylistic takeover (still stable, but assertive)

    If you want โ€œdark but still readableโ€ , start at 1.0 .



    ๐Ÿ”„ Image-to-Image Performance

    YFG Obs performs extremely well in Image-to-Image workflows, especially for:

    • Regrading lighting

    • Darkening bright source material

    • Adding cinematic depth

    • Increasing atmosphere

    • Reinforcing shadow geometry

    Suggested I2I baseline:

    • Denoise: 0.35 โ€“ 0.6

    • Weight: 1.0 โ€“ 1.25


    โš™๏ธ Sampler Recommendation

    Best results observed with:

    โœ… dpmpp_2m_sde_gpu

    This sampler maintains:

    • Smooth gradients

    • Clean shadow rolloff

    • Strong contrast without banding

    • Stable edge definition

    Suggested ranges:

    • Steps: 20โ€“35

    • CFG: 4โ€“6 (adjust with prompt density)


    ๐ŸŒ‘ What YFG Obs Excels At

    • Noir portraits

    • Tactical characters

    • Industrial interiors

    • Dystopian cityscapes

    • Post-apocalyptic themes

    • Moody sci-fi

    • Dramatic stage lighting

    • Subject-in-shadow compositions


    ๐Ÿงช Prompting Tips

    To amplify the effect, add cues like:

    • cinematic lighting

    • low key lighting

    • dramatic shadows

    • volumetric fog

    • rim lighting

    • moody atmosphere

    • industrial setting

    • dystopian city

    To soften the effect:

    • natural lighting

    • bright exposure

    • neutral tone

    • soft daylight

    • and/or lower weight to 0.6โ€“0.8


    ๐ŸŽจ Style Identity

    YFG Obs embodies:

    • Controlled darkness

    • Cinematic intensity

    • Heavy atmosphere

    • Bold tonal contrast

    • Focused subject isolation

    A mood-forward LoRA with high structural stability .


    ๐Ÿ’€ Final Notes

    This is not a lightweight modifier.
    YFG Obs is designed to be bold, confident, and visually assertive โ€”while staying technically stable over a wide weight range.

    If you like dark cinematic compositions with strong tonal control , this one delivers.

    lora
    krea2
    by Sam

    Details

    Downloads
    72
    Platform
    Tungsten
    Platform Status
    Available
    Created
    7/10/2026
    Updated
    7/11/2026
    Deleted
    -

    Files

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.