Many evidences from experimental sources support the description of proteins in solution as an ensemble of conformations in dynamic equilibrium instead of a single structure. Molecular dynamics (MD) and Normal Mode analysis (NM) are two well-established methods to explore conformational landscapes. However, MD is limited by the prohibitive computational cost to reach timescales where large conformational transitions are experimentally observed (from hundreds of μs to ms). NM is well suited to study such motions, but has two main drawbacks: (1) the lack of anharmonicity due to the quadratic approximation of the potential energy surface and (2) the decomposition of motions in independent modes which leads to a non-trivial choice of directions to be studied, especially for large macromolecules. Here, we developed a hybrid approach called AMDNM (Activated Molecular Dynamics in the Normal Modes Space) in which the directions of low-frequency NM are privileged during a short MD. Instead of using constraints or modifying the potential energy function, kinetic energy is regularly injected in the conformational space spanned by the low-frequency normal modes while the MD space is kept in ambient temperature. We developed a protocol to obtain estimates of free energy landscapes based on unbiased MD from clustered conformations generated with AMDNM. We selected two widely studied proteins as test cases: lysozyme and HIV protease (PR). Regarding the conformational space described by the first two low-frequency modes in both proteins, the AMDNM approach outperformed long unbiased MD (up to 1 μs), providing a higher extent of sampling in a few nanoseconds. We also found an excellent correlation with motions inferred from experimental sources (X-ray, EPR and NMR). Finally, our free energy calculations on AMDNM trajectories notably resembled those obtained with other advanced simulation methods such as umbrella sampling and metadynamics, though being less computationally intensive. The obtained conformations may be used in docking studies as well as for the interpretation of experiments such as SAXS curves and NMR.