An arrangement and a method of localizing active speakers in a video conference includes a localization device that locates at least one microphone relative to the camera, while the at least one microphone in turn localizes the relative positions of audio sources. As the microphones are usually positioned close to the audio source in a video conference, the ratio of the distance between microphones relative to the distance between table microphone and audio source is reduced. Thus, the microphones are able to determine the positions of the audio sources with a higher resolution than if placed close to the camera. When the respective positions of the microphones relative to the camera and the audio source are known, the position of the audio source relative to the camera is then determined by means of geometrical calculations.