The charismatic color patterning of organisms has captivated scientists for decades. Animal colors and patterns are important to study as they serve ecologically important functions for communication through sexual or social signaling, and provide crypsis, advertisement, or mimicry. Recent conceptual advances have facilitated the ability to perform a variety of color pattern analyses without needing expensive spectral photography equipment, multispectral images of specimens, nor photoreceptor curves for taxa; hence, we can now use standardized, high fidelity image sets of specimens already accessible to researchers to perform high-throughput studies of color pattern. Yet, many popular color pattern analysis pipelines are halted by requiring users to input the number of dominant color classes (k) before computing color pattern geometry measures. For large-scale analyses, objectively determining k can be cumbersome and inconsistent for groups with wide color class disparity between taxa. Previous studies have avoided this bottleneck by choosing one, overarching k-value to account for the typically observed variation within that group; however, this naive approach fails to account for differences in color pattern geometry statistics when intragroup color diversity is large. Here, we present charisma, an R tool to automatically determine the number of distinct color classes within an image or image set, substantially limiting the need to generalize k for large-scale color pattern analysis. Our toolkit utilizes flexible parameters to yield reproducible, objective, and customizable results for a diverse range of problems, and is designed to work seamlessly with popular color pattern analysis packages (e.g., pavo, patternize).