The density functional theory (DFT) with commonly used furictionals is known to be incorrect for charge-transfer problems. With long-range-corrected (LC) density functionals, the asymptotic exchange potential is gradually switched to the Hartree-Fock exchange at a long range, and the prediction for charge-transfer states is greatly improved. In this work, we test LC-DFT's performance on charge-transfer couplings. The range-separation parameter can be tuned nonempirically for properties of a generalized DFT. We propose to minimize the difference of highest-occupied Kohn-Sham orbital energy and the ionization potential (for hole transfer) or the lowest-unoccupied orbital energy and the electron affinity (for electron transfer). Far photoinduced charge transfer, the minimum in the sum of such differences for the donor and the acceptor is proposed. With the range-separation parameters optimized, we found that ET couplings derived from the LC-DFT are close to those derived from coupled duster with singles and doubles. When compared with experimentally derived Mulliken-Hush couplings, LC-DFT couplings are greatly unproved as well. We also found that the couplings from BNL and LC-BLYP functionals are generally better than those from LC-omega PBE and LC-omega PBE0 LC-DFT is suitable for calculating ET coupling, especially with this nonempirical approach for the range-separation parameter.