In this article, we describe the main methodological steps taken in the InterPhonologie du Français Contemporain (IPFC) research project to build an international corpus of spoken French as a Foreign Language for research and educational purposes. We offer a brief illustration of our approach with a four-session longitudinal study of 12 beginner Japanese learners of French over two years, especially /b/ and /v/ consonants produced in two tasks (repetition and reading of an identical wordlist). Our results illustrate how our perception-based first-screening coding approach provides us with rich information about the developmental profile of the learners, taking into account the task (repetition vs reading), the position in the word (initial, medial, final), the perceptual phonetic characteristics, and ultimately the left and right phonological context of the structure under scrutiny. This work is the first step in an iterative approach to further test experimentally subsets of the non-target-like productions from an acoustic-perceptual point of view.