Find the frame that breaks rank.
Compare palette and rendering density across a whole poster run. The metric finds low-level drift; you decide whether that drift is intentional, symbolic, or worth revising.
No upload leaves your browser. Similarity is not quality, symbolism, provenance, authenticity, or AI detection.
Campaign awaiting pixels
Choose a prediction, then run the bundled sample.
What moved, exactly?
RGB / edgesA flag is a question, not a verdict.
Color means and Canny edges can reveal a campaign mismatch. They cannot tell whether the mismatch is deliberate, whether a render matches a specific animation era, or who made it.
Review next: compare the flagged poster against authenticated current-style references, then inspect composition, symbolism, typography, and asset provenance by eye.
Method: OpenCV RGBA mean, grayscale Canny edges at 70/140, feature normalization, Euclidean distance from the campaign centroid. Small or highly varied sets can make centroid distance unstable.