In real-time towed array systems, performance degradation of array gain occurs when beamforming is carried out on the sensor outputs of a line array which is not straight. In this paper, a new method is proposed for array shape estimation. The procedure consists of two steps. First, the tow-point-induced motion is formulated in the time domain based on the constraints from the tow-point compass-sensor readings and from a discretized Paidoussis equation. At each time instance, the shape estimate is solved from a linear system of equations. It is shown that this solution is equivalent to a previous frequency-domain solution, while the new approach is much simpler. In the second step, the tail compass-sensor data are used to adjust the overall array shape. By noting that variations in the ship speed lead to a distortion in the normalized time axis, the predicted tail displacements are first registered with the tail sensor readings along the time axis. Then distortions in the estimated array shape over its length can be compensated accordingly. A slow-changing bias between sensor zeros is also modeled in order to remove systematic sensor errors. The effectiveness of the new algorithm is demonstrated with real sea-trial data.