In a camera tracking method and apparatus, and a device, a director device determines speech information of speaker in a preset time period. The director device determines a current speaker based on second video information collected by the first camera at a current moment and second audio information collected by the plurality of MICs at the current moment. The director device controls a director status of at least one director camera based on the speech information of speaker, the current speaker, and a speaker shot by the at least one director camera at the current moment. The director camera is configured to send a director video stream to another conference site. This avoids frequent and unnecessary switching of the director camera.