How we verify Anatomical accuracy

Build

29 Oct

Arcanium Studios maintains a high quality control tolerance of 97% or greater. Our artists work with AI to maintain this precision and provide customers the transparency of our methods.

Using a blend of traditional sculpting practices and makeup artist concepts, we deliver sculpts that aim to bear a close resemblance to the subject.

From Claim to Proof: How We Verify Anatomical Accuracy

Facial landmark workflow — professional sculpt reference vs AI × Human (Valehart)

Measurement

Manual side-by-side comparison; sculptor judges likeness visually or with digital overlay.

Five landmark points captured (brow centre, nasal tip, left/right cheilions, chin). RMS deviation auto-calculated.

Adjustment

Sculptor reworks geometry by eye; precision depends on skill and reference lighting.

AI flags outliers and proposes micro-vector shifts (< 0.3 mm); human confirms or rejects visually.

Lighting / Capture Control

Photographs taken under uncontrolled lighting and distance.

Reference and sculpt normalised for scale, angle, and illumination; D65 daylight or calibrated WB.

Acceptance Criteria

Visual “close enough” judgement; minor proportional error tolerated.

RMS ≤ 0.30 mm = pass, 0.31–0.50 mm = review, > 0.50 mm = reject. Variance logged numerically.

Record & Repeatability

Progress photos; little quantitative traceability.

QC log records RMS, landmark set, lighting data, and correction vector per revision.

Why it matters: Both workflows aim for likeness, but the Valehart method quantifies it. RMS (Root Mean Square Deviation) expresses the average landmark variance between a sculpt and its reference in millimetres — letting artistic judgement sit inside measurable tolerance.

METHOD:
RMS = √((Σ Δ²) / n) where Δ = difference in mm between matched landmarks, n = number of points (5).

SAMPLE OUTPUT:
Differences (mm): 0.22, 0.18, 0.27, 0.21, 0.24 → RMS = 0.23 mm ≈ 98 % structural likeness (≤ 0.30 mm threshold).