This paper focuses on instruction-giving practices, a crucial but under-researched aspect of online language tutorials. The context for this qualitative study is a telecollaborative exchange focussing on French as a foreign language. We investigate trainee teachers' instructions for a role-play rehearsal task during webconferencing-supported language teaching sessions. Multimodal (inter)action analysis (Jewitt, Bezemer, & O'Halloran, 2016; Norris, 2004) of the data from three sessions reveals how the trainees mark different stages in the instructions using gaze and webcam proximity, allocate roles helped by the use of gaze (Satar, 2013) and gestures (Guichon & Wigham, 2016; McNeill, 1992), and introduce key vocabulary using word-stress, gaze and text chat strategies. The paper sheds light on the need to demonstrate clear boundaries between instructions and beginning of the task and the need, in future online teacher training programmes, to prepare trainees to direct learners' attention to the resources needed for task accomplishment, explain how the task will be accomplished using the online resources and harness the potential of semiotic resources during this teaching phase.