This randomized parallel-group study examined the effects of curvilinear sprint (CS) training with different radii on linear sprint (LS) performance, CS ability, change of direction (COD), and horizontal forcevelocity (FV) profile in youth soccer players. Eighteen male youth soccer players from the same academy were randomly assigned to a narrow CS group (NCSG; 5.15 m radius) or a wide CS group (WSCG; 11.15 m radius). Both groups completed 12 CS sessions over 6 weeks (2 sessions/week), integrated into their regular team training. Pre- and post-intervention assessments included LS (5–30 m), CS tests on both the good and weak sides over two radii (7.15 m and 9.15 m), modified 505 COD test, and sprint mechanical variables from the FV profile (F0, V0, Pmax, RFmax, Drf). No significant time × group interactions were found for any variable (p > 0.05), indicating that narrow and wide CS training produced similar adaptations. However, within-group analyses revealed significant improvements (p ≤ 0.05) in most CS tests (except for the good side on 9.15 m), in 5 m and 20 m LS times and COD performance. Additionally, the NCSG and WCSG showed small to moderate significant enhancements (p ≤ 0.05) in Pmax and RFmax after the training intervention. In conclusion, 6 weeks of CS training improved LS, CS ability, COD performance, and key sprint mechanical outputs in youth soccer players. These findings suggest that CS training is a specific and effective method for enhancing multidirectional speed and sprint mechanics in this population.