ELSA Speak — AI pronunciation coaching for EFL learners in Vietnam

ELSA (English Language Speech Assistant) was co-founded in Vietnam in 2015 to give EFL learners in Vietnam and across the Global South access to pronunciation coaching they could not otherwise afford, using deep-learning automatic speech recognition to detect errors at the phoneme level without a human teacher.

The app aims to deliver real-time, personalised phoneme-level pronunciation feedback at a cost and scale that native-speaker tutoring cannot match. Several Vietnamese higher-education institutions have formally embedded it in English pronunciation courses in place of, or alongside, one-to-one tutoring.

Learners record spoken English into the app, which applies proprietary ASR to flag phoneme-level pronunciation errors and give instant personalised feedback. The platform has scaled to 195 countries and over 50 million users, with Vietnam historically its largest market (80-90% of early users, now around 20% as the platform globalised).

A peer-reviewed PLOS ONE study (January 2025, n=205 English-major students at a Vietnamese university) found high self-reported satisfaction with usefulness, ease of use and pronunciation improvement. A pre-/post- study at Vien Dong College, Ho Chi Minh City (2024, n=63, one semester) found 77.3% of students rated ELSA effective or highly effective and 81.8% reported increased speaking confidence. A further study in IJIRSS documented high motivation and engagement among Vietnamese EFL learners.

The evidence base is consistent but modest: all published studies rely on self-reported satisfaction and small, non-randomised samples with no independent standardised proficiency measure, and the proprietary ASR model has not been independently audited for accuracy or bias. Data are governed under US CCPA rather than GDPR or Vietnamese law, and systematic integration into formal schooling — rather than self-study — remains the exception.