Human-Centric AI for Southeast Asian Speech and Beyond
MERaLiON builds on A*STAR’s expertise in multimodal AI and speech processing to create a large language model that captures Southeast Asian communication styles. Its architectures capture not only words but also tone, pitch and environmental sounds, tuned to regional patterns such as code switching, local expressions and culturally nuanced emotional cues.
Unlike conventional systems that convert audio into text, MERaLiON processes speech directly, preserving signals such as speaker identity and emotion that are often lost in traditional pipelines. This results in more accurate and natural interactions, with applications across speech translation, call centre analytics, accessibility tools and AI wellness assistants.
Part of Singapore’s $70M National Multimodal Large Language Model Programme, MERaLiON is open source and backed by partners including ST Engineering, Grab, Singtel and Microsoft, driving adoption across the region.
For enquiries and collaboration opportunities, please use our contact form.



