Sound detection with optical means is an appealing research topic. Laser speckle is one effective method to detect the sound due to its sensitivity to tiny motion. The traditional laser-speckle sound recovery methods require a high-speed camera to record fast-moving speckle images and then recover the original sound signals from the motion information of the captured speckle images. In this manuscript, a laser microphone system is proposed to detect and regenerate the sound signal in real time. In the proposed system, only a small part of the imaging sensor is used to ensure a high sampling rate with a common industrial camera and reduce the computation time consumption. Meanwhile, optical flow algorithm is employed to obtain the motion information of captured speckle images and regenerate the sound. These two points allow us to capture images from the camera and regenerate sound in real time without storing any data into the computer, which greatly increases the speed of the system and achieves a microphone-like functions. Experiments are conducted to show that the proposed system can detect and regenerate the sound signal in real-time with a high quality.