We consider the downlink of cell-free massive multiple-input multiple-output (MIMO) systems with orthogonal time frequency space (OTFS) modulation. Two pilot-based channel estimation schemes, namely superimposed pilot-based (SP-CHE) and embedded pilot-based channel estimation (EP-CHE), are applied to estimate the channels at the access points (APs). The SP-CHE scheme superimposes low power pilots onto the data symbols in the delay-Doppler domain to avoid the spectral efficiency (SE) loss due to null guard intervals used in the EP-CHE scheme. In the case of SP-CHE scheme, we consider a max-min fairness optimization problem to jointly optimize the peruser pilot/data power allocation coefficients and per-AP power control coefficients. The complicated non-convex problem is then iteratively solved through two decoupled sub-problems. Moreover, a max-min fairness problem is cast for the EP-CHE scheme, where the optimization variables are the per-AP power control coefficients. Numerical results show that the proposed resource allocation approaches provide at most 42 and 5-fold increase in the 95%-likely per-user SE for the SP-CHE and EP-CHE scheme, respectively, compared with the uniform power control and in correlated shadowing fading channels.