Foreign accent conversion (FAC) aims to transform a speaker's accent while strictly preserving their linguistic content and unique voice identity. Despite its growing importance in language learning and cross-cultural communication, the field suffers from a lack of open-source, reproducible baselines, hindering fair comparisons between emerging models. In this paper, we propose an interpretable and modular cascaded FAC framework built from off-the-shelf automatic speech recognition (ASR), text-to-speech (TTS), and voice conversion (VC) components. We introduce two distinct architectures: a fundamental ASR$\to$TTS$\to$VC pipeline and an enhanced version that incorporates a text-correction (TC) module to mitigate error propagation from the ASR stage. Our system is evaluated using a comprehensive suite of objective metrics—including speech intelligibility, accentedness, naturalness, and speaker similarity—alongside subjective listening tests and real-time factor (RTF) analysis on both CPU and GPU. Results demonstrate that these cascaded pipelines achieve high intelligibility and robust accent transformation while offering practical quality--speed trade-offs for real-world applications. To foster community progress and provide a transparent reference, we have open-sourced our code and configurations\footnote{\url{https://github.com/ChenxiLiu6/Cascaded_FAC}}.