Hacker News August 21, 2026How We Made a Text-to-Speech Model Respond in Sub-50 msRead full article Share: