This abstract discussed the implementation of speech recognition applications to control the on-off electrical device of speech. The work is divided into two modes. Offline mode and online mode. To use voice commands to turn on-off electrical device. Which the words sound as a statement. The speech sounds in offline mode is saved in SOPARE voice and speech mode are saved online in Google Assistant via the microphone. Then process the Raspberry pi 3 board to send data to the wireless transceiver module to control the power on-off. From testing the control system for opening and closing electrical devices with sound it was found that the sound test in the offline mode of the power-on-off control system with sound the accuracy of 96 percent of the voice command and the accuracy of the lamp according to the voice command spoken 100 percent accuracy. Sound tests in the online mode of the power-on-off control system with sound the accuracy of 8 7 percent voice commands and the accuracy of the lamp according to the voice command spoken 100 percent accuracy. The power-on-off control system with sound. Can be used to effectively control the opening and closing of electrical equipment. Only the user speaks the voice commands as needed and can immediately turn on-off the electrical equipment. Helps reduce steps and shorten the time of use making it more convenient