Many principles for the design of multimedia learning materials share the recommendation to facilitate processing. One prominent example is the modality principle, according to which pictures should be presented with auditory rather than visual texts. Research on desirable difficulties, however, indicates that – unlike short-term learning – long-term learning benefits when processing is more demanding and therefore more effortful. In a classroom experiment (Experiment 1) and in a laboratory study (Experiment 2), we tested whether the modality principle serves long-term learning. In a multimedia presentation on the formation of lightning, we varied the text modality (oral vs. written) and the delay between learning and test (retention and transfer performance tested immediately after instruction vs. one week later). In the immediate tests, there was either an auditory advantage (Experiment 1) or no difference (Experiment 2). However, when learning was tested after a delay, the combined processing of written text and animations led to better transfer performance than oral text and animations in both experiments. This suggests that written text presentation serves as a desirable difficulty that supports long-term learning.