A conventional dummy head or a microphone array with a number of microphones has been used to record spatial audio information. Because of its size and weight, however, the conventional dummy head has been used exclusively for the research purpose and a high computational cost is required to deal with the data captured by the microphone array. We propose an audio capture system called 'Miniature head simulator' that consists of a microphone system and a signal processor. Instead of using a conventional dummy head, a small microphone system that consists of three omni-directional microphones is used to capture acoustic signals. The signals are then encoded to a custom data format and transmitted online. The transmitted data can be decoded and reproduced as binaural signals. Since it deals with only three channels of audio stream, data can be processed with relatively low computational cost. By using a miniature head simulator, spatial audio information can be streamed to the browsers or even to the smartphones of the end-users. It can be used in a remote acoustic sensor and remote control system, sound detection system, video conference, and various other fields of industry. © 2013 Acoustical Society of America.