Speech-to-Text AI for Improving English Pronunciation in ESL Learners

English as Second Language (ESL) learners often struggle with pronunciation, which can hinder academic success and social integration. This study investigates the effectiveness of a speech-to-text artificial intelligence (AI) system in improving pronunciation accuracy among ESL learners. Using a phoneme-matching approach, the system provided real- time corrective feedback to students in semi-urban learning environments. Data were collected through pre- and post-tests measuring accuracy, precision, recall, and F1-score. Results revealed a 15% improvement in pronunciation accuracy, supported by consistent gains across all performance metrics. Learners also demonstrated increased confidence and sustained engagement, highlighting the motivational value of instant AI-based feedback. These findings suggest that speech- to-text AI can complement traditional instruction by offering personalized and continuous pronunciation training. Future research should explore long-term retention and integration with immersive technologies such as virtual and augmented reality.