In healthcare, clinicians are expected to communicate in a compassionate way with patients.1 Although medical training for students at the University of Bern offers communication training, for example, breaking bad news or motivational interviewing, compassion or compassionate behaviour is not yet specifically addressed. An interdisciplinary team (medical education, psychology, communication and palliative care) designed and piloted an elective blended course with 1 day of classroom training to improve sixth-year medical students' compassion in patient–physician communication. The learning goals of the pilot included (a) understanding the role of compassion in the patient-physician relationship; (b) characterise the components of compassion; (c) reflecting one's own emotions and developing self-compassion techniques; (d) demonstrate understanding a patient's situation; and (e) developing a compassionate care plan together with the patient. In order to learn about the existing concepts and the importance of compassion, students (n = 8) received preparatory e-learning material in advance. Its contents were discussed and reflected upon during the classroom training. Preparatory reading included directions regarding self-compassion techniques and two case descriptions to initiate reflection about one's own emotions. These materials were contemplated during the training with an expert input about self-compassion and group discussions about integrating these techniques into daily clinical practice. Using simulation, students then explored their feelings and emotions from a patient perspective through self-exposure to various life conditions, such as physical restrictions (wearing bariatric suits or age simulator) or having stigmatising skin diseases (using a moulage of psoriasis or acne). The goal was to foster a deeper understanding of the patient's situation through own experiences. Based on these experiences (‘how did I feel myself, and what might be the related feeling of my patient?’), students created a communication guide for a conversation with simulated patients (SP). In a second step, students used this guide in SP interactions trying to shift from medical reasoning and counselling to assessing and addressing emotions and individual suffering. The final step in this interaction then was to create a compassionate care plan yielding to integrate medical treatment and emotional comfort. The evaluation of the course included (a) students' verbal feedback immediately after the training, (b) students' written evaluation in respect of content and learning goals as well as methodological and didactic aspects and (c) students' self-assessment of skills in compassionate communication using the Sinclair Compassion Questionnaire (SCQ) prior and after the training (www.compassionmeasure.com). Our interactive compassion training seems to address students' needs and expectations in terms of a high quality and compassionate patient-physician interaction. Students very much appreciated the exchange and the reflections within the group about compassion, communication and their own experiences during simulation. Before the training, the majority rated themselves according to the SCQ in most items as ‘somewhat competent’ or ‘neutral’. After the training, students' ratings were more differentiated, ranging from ‘somewhat lacking competence’ to ‘very competent’. All participants strongly recommended offering this ‘compassion training’ to all medical students as a mandatory class; one student even regarded this training as being one of the most important aspects of medical school. Andrea Lörwald, Seraina Petra Lerch, Daniel Bauer, Steffen Eychmüller and Sibylle Jeanine Felber made substantial contributions to conception and design of the course. Daniel Bauer, Steffen Eychmüller, Sibylle Jeanine Felber, Seraina Petra Lerch, Florence Liaudet and Andrea Lörwald conducted the training and acquired the data for evaluation. Florence Liaudet, Sibylle Jeanine Felber, Steffen Eychmüller, Seraina Petra Lerch and Andrea Lörwald analysed and interpreted the data. Sibylle Jeanine Felber and Andrea Lörwald drafted the manuscript. Daniel Bauer, Steffen Eychmüller and Seraina Petra Lerch revised it critically for important intellectual content. All authors approved the version to be published and agree to be accountable for all aspects of this study.