Time and space constraints in foreign/second language (L2) instruction restrict learners’ exposure to varied speech (Collins & Muñoz, 2016), essential for pronunciation development (Thomson, 2018). High-Variability Phonetic Training (HVPT) offers a promising solution by exposing learners to phonetic variation; however, its implementation remains underexplored (Barriuso & Hayes-Harb, 2018). This study investigates the integration of Text-To-Speech (TTS) technology with HVPT to provide varied L2 input in a semi-autonomous (beyond-the-classroom) environment. This mixed-method study employed a pretest-posttest design to examine discrete and holistic aspects of pronunciation development. The discrete analysis assessed learners’ phonological awareness of regular past tense marking (-ed: walk/t/, play/d/, visit/ɪd/), which aligns with Celce-Murcia et al.’s (2010) initial stage of pronunciation development. In the holistic analysis, eleven English-speaking raters evaluated broader pronunciation aspects (i.e., comprehensibility, accentedness), following Munro and Derwing (1995). Thirty lower-intermediate ESL learners were divided into a Treatment Group (exposed to varied TTS voices) and a Control Group (exposed to a single TTS voice), engaging in ten self-paced activities over four weeks. Results revealed significant improvements in phonological awareness of past -ed allomorphy for both groups, with no statistical difference between them. However, the Treatment Group achieved statistically significant improvements in both comprehensibility and accentedness compared to the Control Group. These findings suggest that although TTS improves phonological awareness irrespective of HVPT implementation, TTS-based HVPT leads to superior pronunciation gains in the holistic measures adopted. This research highlights TTS’s potential to provide varied aural input, enhancing L2 pronunciation and offering insights for developing accessible language learning resources.