This paper deals with a specific type of lip-sync videos on TikTok, in which the spoken text from the audiotrack is combined with divergent written text inserts and thus overwritten. With recourse to the concept of canvas from multimodality theory and to conceptual metaphor theory, it is argued that the overwritings, in which other sign modalities such as facial expressions and gestures also participate, trigger metaphorical readings. Sample analyses show that the metaphorical mapping is directional and always runs from the spoken text as source to the written text as target. The selection of the features relevant for the projection is also guided by gestural and mimic signals, so that a broad variety of semiotic modalities are involved in the metaphorisation process. In addition, the fluid transitions between metaphors on the one hand and similes on the other become clear. Finally, it is discussed how the seriality of memes shapes the metaphorisations, which conversely proves to be the driver and mode of memefication on TikTok.