Statistical tropical cyclone rainfall (TCR) models based on historical observations have the potential to support flood hazard assessment by efficiently generating rainfall fields for large ensembles of TC tracks. Despite efforts to improve the characterization of TCR, existing models exhibit limited performance. A TCR error modeling framework offers an alternative approach to bypass this limitation by characterizing errors through a multiplicative error model comprising deterministic and stochastic components. Here, using the Interagency Performance Evaluation Task Force (IPET) rainfall analysis model as the base model, we apply this framework to 113 local domains along the U.S. Atlantic and Gulf Coasts to develop an error-correction-based probabilistic TCR generator. For the deterministic component, we fit overall bias and rain-dependent bias models to correct systematic biases in the IPET model’s rainfall estimates. The remaining residuals are then modeled probabilistically by characterizing their marginal distribution and spatial dependence. We show that systematic biases in total IPET rainfall are effectively corrected, with the average overall bias increasing from 0.7 to 1.07, close to the unbiased value of 1. We also show that the TCR generator produces spatially plausible rainfall fields and captures the variability of radial rainfall profiles, with reference rainfall intensities falling within the simulated 90% range in 82% of radial bins. This framework is general and can be transferred to other areas of the world and applied to other TCR models and to both observed and simulated TC tracks.