Within the framework of the key approach from the theory of dynamic inversion, input reconstruction problems for stochastic differential equations are investigated. Different types of input information are used for the simultaneous reconstruction of disturbances in both the deterministic and stochastic terms of the equations. Feasible solving algorithms are designed; estimates of their convergence rates are derived. An empirical procedure adapting an algorithm to a specific system’s dynamics to obtain best approximation results is discussed. An illustrative example for this technique is presented.