Langcraft is a research project exploring how speech is actually produced, not just how it is transcribed. Most speech technologies are optimized for converting speech into text, which often hides important differences in pronunciation. Langcraft instead analyzes speech at the acoustic and articulatory level to understand how sounds are formed and where deviations occur. The goal is to build a diagnostic layer for speech, enabling applications such as language learning, accent training, and speech therapy to provide more precise feedback on pronunciation.