This pilot study explores the impact of visualizing real-time speech on improving lexical stress among non-native English speakers. The study introduces a real-time 3D spectrogram, where learners can see, hear and imitate a model's speech and view their own productions. Six French English learners participated in a three-phase within-subject study consisting in a pre-test, a 10-week training session using the spectrogram, and a post-test. The study questions whether visualizing speech improves lexical stress production and equips learners to handle new words post-training. An auditory analysis of pre-test and post-test results revealed a slight improvement in correct lexical stress placement, with the global mean of accurately pronounced words rising from 4 in the pre-test to 4.5 participants in the post-test. The mean for correctly pronouncing pre-test words included in the post-test, improved by 1 (from 4 to 5) but there was minimal improvement in correctly pronouncing the new words in the post test. The goal of this study is to contribute to understanding how L2 learners can improve their word stress accuracy in English and to expand our knowledge regarding multi-sensory tools' efficacy in second language learning.