Parameters of Simulations

SkinGPT offers a comprehensive set of parameters that can simulate various skin effects including environmental impacts, skincare product application, aging, and specific skin parameter changes. The system supports simulations for multiple parameters including:

  • Redness

  • Pigmentation

  • Uniformness

  • Acne

  • Lines

  • Skintone

  • Face Geometry

  • Aging

  • Skin Aging (option of aging parameter)

  • UV Index (option of aging parameter)

  • Weight (option of aging parameter)

  • Hair graying (option of aging parameter)

Most of these parameters can be combined.

List of Parameters

Redness

  • Supported in SkinGPT and SkinGPT High Resolution models.

  • Can be configured to show positive and negative effect.

  • Positive adjustment calms erythema, yielding a more even complexion that reflects reduced inflammation. Negative adjustment heightens flushing to simulate flare-ups typical of rosacea or irritation.

  • Can be configured for:

    • Whole face.

    • T-zone (forehead, nose, and chin).

    • U-zone (cheeks and jawline).

Pigmentation

  • Supported in SkinGPT and SkinGPT High Resolution models.

  • Can be configured to show positive and negative effect.

  • Positive lightens discoloured areas, portraying a clearer, more uniform tone after brightening care. Negative deepens spots and patches, replicating sun damage or hormonal hyper-pigmentation.

  • Can be configured for:

    • Whole face.

    • T-zone (forehead, nose, and chin).

    • U-zone (cheeks and jawline).

Sun Spots

  • Supported in SkinGPT and SkinGPT High Resolution models.

  • Can be configured to show positive effect only.

  • Positive lightens solar lentigines and mottled pigmentation, yielding a clearer, more even complexion that suggests successful photodamage reversal. Negative darkens and multiplies sun spots, conveying cumulative UV exposure or inadequate protection.

  • Can be configured for:

    • Whole face.

    • Forehead.

    • Left cheek.

    • Right cheeks.

    • Nose.

    • Perioral.

Uniformness

  • Supported in SkinGPT and SkinGPT High Resolution models.

  • Can be configured to show positive and negative effect.

  • Positive smooths colour transitions for a balanced, consistent skin surface. Negative introduces blotchiness and visible tone variation, highlighting uneven texture or PIH.

  • Can be configured for:

    • Whole face.

    • T-zone (forehead, nose, and chin).

    • U-zone (cheeks and jawline).

Acne

  • Supported in SkinGPT model only. Not supported in SkinGPT High Resolution model.

  • Can be configured to show positive and negative effect.

  • Positive removes comedones and pustules, illustrating clearer skin following effective therapy. Negative adds lesions to depict breakout severity or product misuse.

  • Can be configured for:

    • Whole face.

    • T-zone (forehead, nose, and chin).

    • U-zone (cheeks and jawline).

Lines

  • Supported in SkinGPT and SkinGPT High Resolution models.

  • Can be configured to show positive effect only.

  • Positive softens fine lines and wrinkles for a rejuvenated appearance. Negative accentuates creases and furrows, conveying progressive intrinsic ageing or dehydration.

  • Can be configured for:

    • Whole face.

    • T-zone (forehead, nose, and chin).

    • U-zone (cheeks and jawline).

Skintone

  • Supported in SkinGPT and SkinGPT High Resolution models.

  • Can be configured to show positive and negative effect.

  • Positive brightens and balances the base complexion, reducing dullness and unwanted undertones. Negative introduces sallowness or uneven hue to showcase fatigue or environmental stress.

  • Can be configured for:

    • Whole face.

    • T-zone (forehead, nose, and chin).

    • U-zone (cheeks and jawline).

Face Geometry

  • Supported in SkinGPT model only. Not supported in SkinGPT High Resolution model.

  • Positive restores volume and firmness, subtly lifting contours to suggest youthful plumpness. Negative models volume loss and sagging, simulating gravitational ageing or weight change.

  • Can be configured for:

    • Lips

    • Nasolabial Folds

    • Eyebrows

    • Jaw

Weight

  • Supported in SkinGPT model only. Not supported in SkinGPT High Resolution model.

  • Positive adjustments create a slimmer look. Negative adjustments add weight to the face.

  • Works only when the Aging parameter is enabled. Does not work when the Skin Aging is turned on.

  • Supported range

    • From -100 to 100 in conditional units.

Hair graying

  • Supported in SkinGPT model only. Not supported in SkinGPT High Resolution model.

  • Adjustments cause hair to appear grayer.

  • Works only when the Aging parameter is enabled. Does not work when the Skin Aging is turned on.

  • Supported range

    • From -100 to 0 in conditional units.

Aging

  • Supported in SkinGPT model only. Not supported in SkinGPT High Resolution model.

  • Can be configured for:

    • Age Shift

      • Adjusts the perceived age relative to the current age.

      • Example: A shift of +10 will age the face by 10 years; -5 will rejuvenate it by 5 years.

    • Specific Age

      • Directly sets the target age for visualization.

      • Example: Setting age to 45 will simulate the person looking approximately 45 years old, regardless of their current age.

  • Positive reverses multiple age markers—fewer wrinkles, spots, and laxity—producing a younger look. Negative advances these markers in concert to portray accelerated ageing.

  • Supported age range

    • From 20 to 70 years old.

    • Values outside of this range may not be supported or could produce inaccurate results.

Skin Aging

  • Allows aging to be applied to the facial skin only, without changing basic personality traits such as facial shape, eye or hair color.

  • Supported in SkinGPT model only. Not supported in SkinGPT High Resolution model.

  • Skin Aging is an option of Aging parameter and work with Age Shift only.

UV-index

  • Supported in SkinGPT model only. Not supported in SkinGPT High Resolution model.

  • Works only when the Aging parameter is enabled.

  • When UV Index is turned on, the Pigmentation parameter will be ignored, as pigmentation changes are derived from the UV simulation logic.

  • Accepts values from 0 to 11+, following the global UV index scale.

    • 0–2: Low UV exposure

    • 3–5: Moderate exposure

    • 6–7: High exposure

    • 8–10: Very high exposure

    • 11+: Extreme exposure

Positive and negative effect explanation

SkinGPT uses a relative scale from -100 to +100 to simulate changes in skin conditions. Here’s how to interpret and apply this scale across supported parameters.

Key Principles

  • 0: Represents the user’s original, unaltered state.

  • +100: Simulates a strong improvement in the condition — enough to move the result into the “problem-free” group based on the underlying metric distribution.

  • –100: Simulates a strong worsening of the condition — enough to place the result into the “severe issue” group.

Relative vs Absolute Change

Important: The effect of a +100 or -100 adjustment does not translate to an exact numeric shift in the metric (e.g., not always +20 or –20 points). The outcome depends on the initial distribution of that metric in the dataset.

  • Each skin metric (e.g., redness, pigmentation, lines) has its own distribution curve (often centered around a mean like 60–70).

  • Applying +100 will push the simulated result into the upper quantile (best visible outcome), and –100 into the lower quantile (worst visible outcome).

Visual Analogy

Think of each metric as a bell curve:

  • +100 shifts the individual to the right end of the curve (least concern).

  • –100 shifts them to the left (highest concern).

  • The same shift amount can produce different absolute outcomes depending on where the person starts on the curve.

Notable Clarifications

  • A person with moderate redness and another with severe redness will both experience improvement with +100, but the final result may not be identical.

  • The scale is normalized per metric and designed to be intuitive, but not linear in effect.

Exception Cases

Some metrics (e.g., Sunspots, ITA (Individual Typology Angle)) use a linear scale and do not follow the distribution-based transformation model described above.

Last updated

Was this helpful?