A method and a system for supporting medical personnel in a procedure on a patient are provided. Medical image data of the patient and of a medical object is continuously captured. Using the image data, a digital patient model is generated and continuously updated, and a position of the medical object is continuously tracked. Through automatic processing of the image data, a situational and/or spatial context in which the medical object is situated is determined. Based on the determined context from the patient model, speech data that describes at least part of the determined context in speech form is automatically generated. The speech data is then output to the medical personnel by an acoustic speech output and/or as text on a display surface. Thus, the medical personnel may be informed particularly reliably and with little distraction about the respective current situation.