The void ratio limits and their range are mostly governed by their grain size, distribution, particle shape and fabric for cohesionless soils. For the purpose of developing a probability-based predictive model, a database, which consists of minimum and maximum void ratios, their range, as well as particle morphology (roundness, sphericity), size and distribution (mean grain size, fines content and coefficient of uniformity) parameters, is compiled. A set of predictive models, which use the combinations of C-u, D-50, FC, R, and S parameters, is proposed. The models with C-u and/or D-50, R and S, produce the most precise predictions. The probabilistic forward use of the proposed models is illustrated by a practical example. They are also presented in the form of deterministic chart solutions, illustrating the variation of in-situ void ratios for varying density states as functions of uniformity coefficient or mean grain size.