This study investigated the interaction between emotion-eliciting pictures and power output during a repetitive supra-maximal task on a cycle ergometre. Twelve male participants (mean (±SD) age, height and weight: 28.58 ± 3.23 years, 1.78 ± 0.05 m and 82.41 ± 13.29 kg) performed 5 repeated sprint tests on a cycle ergometre in front of neutral, pleasant or unpleasant pictures. For each sprint, mechanical (peak power and work), physiological (heart rate) and perceptual (affective load) indices were analysed. Affective load was calculated from the ratings of perceived exertion, which reflected the amount of pleasant and unpleasant responses experienced during exercise. The results showed that peak power, work and heart rate values were significantly lower (P < 0.05) for unpleasant pictures (9.18 ± 0.20 W ∙ kg−1; 47.69 ± 1.08 J ∙ kg−1; 152 ± 4 bpm) when compared with pleasant ones (9.50 ± 0.20 W ∙ kg−1; 50.11 ± 0.11 J ∙ kg−1; 156 ± 3 bpm). Furthermore, the affective load was found to be similar for the pleasant and unpleasant sessions. All together, these results suggested that the ability to produce maximal power output depended on whether the emotional context was pleasant or unpleasant. The fact that the power output was lower in the unpleasant versus pleasant session could reflect a regulatory process aimed at maintaining a similar level of affective load for both sessions.