Background: Comfortable gait speed after stroke is key to community ambulation, but the roles of step/stride-length mechanics, balance capacity, and balance confidence remain uncertain. Objects: To identify independent determinants of comfortable gait speed in ambulatory adults with chronic stroke and to clarify the roles of spatiotemporal gait mechanics and clinical balance measures. Methods: A cross-sectional secondary analysis was conducted in 40 inpatients classified as Functional Ambulation Category level 4. Comfortable overground speed from the 10-Meter Walk Test (10MWT) was the outcome. Candidate determinants were stride length, paretic step length, paretic single-limb support, Berg Balance Scale (BBS), limits-of-stability (LOS) area, center of pressure (COP) sway velocity, and Activities-specific Balance Confidence (ABC). Pearson correlations with Benjamini–Hochberg false discovery rate control were primary; Spearman’s ρ and partial correlations (adjusting for stride and paretic step length) tested robustness. Results: Speed was strongly correlated with stride length (r = 0.954) and paretic step length (r = 0.941), with balance capacity (BBS r = 0.916; LOS r = 0.868), and fell as sway increased (COP r = −0.878; all q < 0.001). Single-limb support showed a moderate link (r = 0.734), while ABC was essentially null. After accounting for stride and paretic step length, BBS, LOS, and COP still mattered; single-limb support and ABC did not. A sensitivity analysis using 10MWT time (s) as the outcome produced associations of similar magnitude: stride length and BBS, which were positively associated with gait speed, were negatively associated with time, whereas COP sway, which was negatively associated with gait speed, was positively associated with time. Conclusion: In ambulatory adults with chronic stroke, comfortable gait speed appears to be driven primarily by step/stride-length mechanics and objective balance capacity, whereas balance confidence did not add explanatory value in this cohort. Speed-focused rehabilitation may need to prioritize both step-length mechanics and balance capacity, while the role of confidence appears limited in this dataset.