This study evaluated a procedure for estimating in vivo Achilles tendon (AT) force from ultrasound images. Two aspects of the procedure were tested: (i) accounting for subject-specific AT stiffness and (ii) accounting for changes in the relative electromyographic (EMG) intensities of the three triceps surae muscles. Ten cyclists pedaled at 80 r.p.m. while a comprehensive set of kinematic, kinetic, EMG, and ultrasound data were collected. Subjects were tested at four crank loads, ranging from 14 to 44 N·m (115 to 370 W). AT forces during cycling were estimated from AT length changes and from AT stiffness, which we derived for each subject from ultrasound data and from plantar flexion torques measured during isometric tests. AT length changes were measured by tracking the muscle-tendon junction of the medial gastrocnemius (MG) relative to its insertion on the calcaneus. Because the relative EMG intensities of the triceps surae muscles varied with load during cycling, we divided subjects’ measured AT length changes by a scale factor, defined as the square root of the relative EMG intensity of the MG, weighted by the fractional physiological cross-sectional areas of the three muscles, to estimate force. Subjects’ estimated AT forces during cycling increased with load (p<0.05). On average, peak forces ranged from 920±96 N (14 N·m, 115 W) to 1510±129 N (44 N·m, 370 W). For most subjects, ankle moments derived from the ultrasound-based AT strains were 5 to 12% less than the net ankle moments calculated from inverse dynamics (r2=0.71±0.28, RMSE=8.1±0.33 N·m). Differences in the moments increased substantially when we did not account for changes in the muscles’ relative EMG intensities with load or, in some subjects, when we used an average stiffness, rather than a subject-specific value. The proposed methods offer a non-invasive approach for studying in vivo muscle-tendon mechanics.