Since the powertrain systems of electric vehicles (EVs) lack the traditional engine sound, their NVH performance differs from that of conventional fuel-powered vehicles, making the use of active sound design (ASD) systems increasingly common to provide compensatory sound. With the increasing demand for ASD systems, sci-fi sounds are emerging as a design direction to enhance the acoustic feedback of powertrain systems and to elevate the futuristic and immersive driving experience of vehicles. A method for generating sci-fi soundscapes using a granular synthesis algorithm is proposed in this paper. First, based on the designed sci-fi target sound characteristics, a sound grain generation method using the adaptive principal frequency technique is proposed, and the overlap-and-add (OLA) method is employed to synthesize the sound grains. Then, to enhance the sound continuity and smoothness during the sound synthesis process, a method for optimizing a composite cosine window function using a genetic algorithm is proposed. Finally, to verify the effectiveness of the proposed method, simulation and subjective evaluation experiments are conducted. The results indicate that the sound synthesis algorithm effectively reproduces the target sci-fi sound, and the composite cosine window function, optimized by a genetic algorithm, enhances the auditory smoothness of the synthesized sound grains. This approach offers a solution for powertrain sound compensation, which plays an important role in improving the interior NVH performance of EVs.