Hybrid brain-computer interfaces (BCIs) have been proved to be more effective in mental control. In this study, a hybrid BCI speller system combining steady-state visual evoked potentials (SSVEPs) and eye tracking has been proposed. In this system, the eye tracker was used to detect eye gaze position for a 3×3 block selection, after that classification of the command was achieved through filter bank canonical correlation analysis (FBCCA) method. Results showed that the 40-classes hybrid speller system outperformed the SSVEP-only method, achieved a mean accuracy of 92.1% and a mean information transfer rate (ITR) of 180.8 bits/min during online experiments, and the scalability of the proposed system also has been tested with larger number of commands.