X-ray computed tomography with laboratory imaging chains often struggles with high-speed processes, as recording a single tomographic dataset quickly enough is often a challenging task. This paper presents a method for extracting the eigenmode of a harmonically excited oscillating object based on a probabilistic analysis of its tomographic reconstruction. In the standard reconstruction of an oscillating object, where the recording of tomography data is realised over a relatively long period of time, the highest probability of the object occurrence is in its amplitudes. Based on this fact, it is possible to identify the eigenshape of the oscillating object by searching for the envelope of its motion. The identified modal shapes show good agreement with the laser Doppler vibrometer measurements. Consequently, the effectiveness of the method was demonstrated for objects that are unsuitable for traditional laser vibrometry due to their shape or surface limitations.