How can an editable phoneme representation remain explicit through local speech synthesis?
IPA-core structures, validation, edited-phoneme synthesis and a loopback browser UI are implemented. English is the current MVP boundary; timing highlights are estimated, not forced alignment.