As larger particles require more energy for transport, grain size is a key indicator for the magnitude of geohazards and other depositional processes. However, sample size requirements and laborious laboratory procedures limit our ability to extract this information at human-relevant (years to decades) timescales. The emergence of non-destructive and high-resolution core scanning techniques offer a solution to upscale measurements by mapping grain size-sensitive parameters at μm instead of cm scales. These include X-Ray Fluorescence (XRF) – tracking variations in elemental geochemistry that are often linked to mineral grain size, and Computed Tomography (CT) – capturing differences in density that control size sorting during deposition. Recent work demonstrates that these relations can be captured with linear regression fits, thus paving the way for predictive grain-size modelling approaches. Here, we expand on this work by assessing the potential of CT greyscale density data as a predictor. To do so, we developed a controlled experiment using synthetic sediment records (phantoms) – varying grain size, geochemistry, as well as two major sources of noise: organic and water content. Our results show that CT data can be used as a sole predictor, especially in cases where a homogenous mineralogy limits the use of XRF geochemistry-based approaches. Applications on natural sediment cores containing reworked volcaniclastics confirm these findings under real-world conditions, and highlight the complementarity of CT and XRF data. Finally, we present a code-free web-based workflow to make the presented grain-size prediction approach readily accessible to the wider geoscience community.