In this paper, an adaptive prescribed performance control scheme is proposed based on time-varying barrier Lyapunov function for flexible-joint manipulator systems with output constraints and model uncertainties. A time-varying tangent barrier Lyapunov function is first presented by constructing a time-varying constrained boundary which attenuates exponentially, and it extends the application scope of the conventional logarithmic barrier Lyapunov functions. In addition, by presetting the parameters of the time-varying boundary function, the system output has the smaller overshoot and faster tracking speed in the initial stage, and the satisfactory steady-state performance can be guaranteed simultaneously. Then, the feedback control law is designed by employing the backstepping technique to ensure the output constraints and the trajectory tracking accuracy. All the closed-loop signals are proved to be uniformly ultimately bounded through using the Lyapunov stability theorem, and numerical simulations are given to show the effectiveness of the proposed scheme.