Among the most important atmospheric processes to humans is precipitation, which may take the liquid phase (rainfall) or ice phase (snowfall) at the Earth's surface. However, the great majority of precipitation reaching the Earth's surface passes through an ice phase before melting, and thus descends some distance through the atmosphere at a rate that is commonly understood to depend on ice particle shape. While it is colloquially said that no two snowflakes are exactly alike, their shapes do fall into a range of categories. In this work, a common diversity of ice crystal shapes are reproduced via 3D printing and their shapes are found to lead to a range of stable and unstable patterns of motion, such as zigzagging or spiraling. These motions are systematically investigated and characterized. Such advances in understanding the variability of ice fall speeds bear on a wide range of disciplines including climate forecasting and a variety of approaches to remote sensing of atmospheric conditions. [Videos are recommended accompaniment.]