Investigating muscle architecture in static and dynamic conditions is essential to understand muscle function and muscle adaptations. Muscle architecture analysis, primarily through extended field-of-view ultrasound imaging, offers high reliability at rest but faces limitations during dynamic conditions. Traditional methods often involve “best fitting” straight lines to track muscle fascicles, leading to possible errors, especially with longer fascicles or those with nonlinear paths. Moreover, muscle architecture varies along the same muscle, with potential differences in curvature. This study aimed to develop and test a new software for muscle architecture characterization considering fascicle curvature during dynamic conditions. Muscle architecture data from different muscle regions using various digitalization methods were compared. Ten healthy young adults (24.1 ± 1.6 years;​ 177.7 ± 7.4 cm;​ 72.7 ± 7.7 kg;​ 9M/1F) performed maximal knee extension at 75°.s−1 while B-mode ultrasound images of vastus lateralis muscle were captured in two muscle sites (at 50 % and 83 % of femur length). The analysis involved automated straight-line (ST) methods and custom manual linear extrapolation (MLE) software with segmented fascicle tracking using 2 (MLE2) and 4 (MLE4) segments inside the field of view. Results indicated significant overestimations of fascicle length, muscle belly length and thickness and underestimation of pennation angle using ST compared to MLE methods, especially in the distal region. Intra-rater repeatability for MLE4 was excellent (ICC = 0.93;​ 0.90;​ 0.93;​ 0.88, respectively;​ P < 0.001), while inter-rater reliability varied. This study confirms the need to consider fascicle curvature for accurate resting muscle architecture characterization, even in the middle region of the muscle, and extends these considerations to dynamic conditions.