Mobile application performance is crucial for enhancing product competitiveness in the market. As mobile application development grows more intricate, automated performance testing has become increasingly challenging. Most testing tools rely on source code or binary files, which limits the evaluated scenarios and requires testers to possess coding skills. This paper presents a full-process intelligent speed performance testing tool, SmartPerf, from automated test case recording to analysis report output. SmartPerf simulates user behaviors to record cross-platform test scripts. Subsequently, the tool segments the recorded screen data into frames during replaying, and the general transition point recognition automatically identifies the first and last frames. SmartPerf also supports uploading long screen recording videos and determines each transition point's first and last frames in continuous multiple actions switching through the temporal convolution segmentation. This paper introduces the user-friendly evaluation metric, mAcc@5. The results indicate an accuracy of 0.9 for general transition point recognition in single actions and 0.76 for temporal convolution segmentation in continuous multiple actions, significantly reducing labor costs. The conclusions given by the tool reports reflect the performance quality effectively, consistent with the market reality.