Multivideo summarization aims to fuze multiple videos of sightseeing into a condensed, descriptive, and aesthetically pleasing one. It is a useful technique that can be applied for video displaying, storage, and communication. In this paper, we propose a new summarization technique that enforces video stability and preserves well-aesthetic frames. In particular, a multitask feature selection is introduced to efficiently discover the semantically important features. Then, the key frames are selected based on their contributions to reconstruct the video semantics. Thereafter, a probabilistic model is proposed to dynamically fit the key frames into an aesthetically pleasing video summary, wherein its constituent frames are adaptively destabilized. Experimental results on multiple videos taken from prominent sceneries throughout the world demonstrate the descriptiveness, aesthetics, and stability of the generated summary.