Many species produce rhythmic sound sequences. Some purportedly speed up their vocalizations throughout a display, reminiscent of - but not necessarily equivalent to - \"accelerando\" in human music. This phenomenon has been frequently reported but rarely quantified, which limits our ability to understand its mechanism, function, and evolution. Here, we use a suite of rhythm analyses to quantify temporal and acoustic features in the display songs of male African penguins (Spheniscus demersus). We show that songs get faster (i.e., accelerando) and louder (i.e., crescendo) as they progress. The accelerando occurs because the inter-syllable silences, not the syllables themselves, predictably shorten over time. This rhythmicity is maintained even when individuals take audible breaths. Individuals also show plasticity: when they start with a slow tempo, they speed up more strongly than when they start with a fast tempo. We hypothesize that this well-timed accelerando may stem from arousal-based mechanisms, biomechanical constraints, or more complex rhythmic control; future work should test the mechanisms behind this intra-individual rhythmic variation, since non-passerine birds are thought to have limited vocal plasticity. By integrating a rich empirical dataset with cutting-edge rhythm analyses, we establish the necessary foundation to determine how such features evolved and their role(s) across communication systems.