Protect the few cues that make the character unmistakable
Anime faces can be visually simple and still be extremely identity-sensitive. Check every reference against three tiers in order. Tier 1, identity landmarks that must survive: eye shape and spacing, brows, mouth direction, face framing, hairline, mask edges, and one or two signature features. Tier 2, silhouette and separation: outer contour, hair mass direction, and the open channels that keep Tier 1 separable. Tier 3, negotiable support: hair strand counts, costume folds, ornament density, micro-texture, repeated hatching and tiny accents.
Run keep, simplify, drop triage across the tiers: keep Tier 1 at full priority, simplify Tier 2 repeated structure (reduce hair to mass plus a few form strands, thin folds to direction, cut ornament counts), and drop Tier 3 first when the tightest gap or densest cluster cannot stay open at the planned width. For mask-led references the Tier 1 pair is mask ears plus the eye-muzzle map; for prop-led portraits (katana, staff, helmet, compass, visor) the Tier 1 pair is face plus prop silhouette, with costume and tech detail in Tier 3.
Failure follows a pattern across the twelve examples grouped into five treatments. The character portrait loses its smile and hairline when costume detail is kept at equal weight; the portrait-plus-mask build loses both faces when ornament competes with either one; the mask-like portrait loses expression when teeth, visor reflections and facial texture are preserved instead of the eye and mouth channels; the graphic clean build loses its forgiving read when added decoration fills its open channels; the creature example loses its read when small accents bridge the gaps between face, limbs and contour. When any of those failures appear at true size, simplify the failing cluster first and recheck before redrawing anything structural.













