We have developed a source-receiver compression approach for reducing the computational time and memory usage of the acoustic and elastic full-waveform inversions. By detecting and quantifying the extent of redundancy in the data, we assembled a reduced set of simultaneous sources and receivers that are weighted sums of the physical sources and receivers used in the survey. Because the numbers of these simultaneous sources and receivers could be significantly less than those of the physical sources and receivers, the computational time and memory usage of any gradient-type inversion method such as steepest descent, nonlinear conjugate gradient, contrast-source inversion, and quasi-Newton methods could be reduced. The scheme is based on decomposing the data into their principal components using a singular-value decomposition approach, and the data reduction is done through the elimination of the small eigenvalues. Consequently, this would suppress the effect of noise in the data. Moreover, taking advantage of the redundancy in the data, this compression scheme effectively stacks the redundant data, resulting in an increased signal-to-noise ratio. For demonstration of the concept, we produced inversion results for the 2D acoustic Marmousi and BP models for surface measurements and an elastic model for crosswell measurements. We found that this approach has the potential to significantly reduce computational time and memory usage of the Gauss-Newton method by 1–2 orders of magnitude.