Speech processing, the science of teaching computers to understand, interpret and generate human speech, has become one of the fastest-moving fields of artificial intelligence and underpins technologies millions use daily, from asking Siri for directions and dictating text messages to turning on live captions in online meetings and asking AI assistants questions out loud. As thousands of researchers prepare to gather in Sydney for Interspeech 2026, the world's leading conference on speech science and speech technology, UNSW Associate Professor Beena Ahmed explains that devices break audio into small segments starting with the sound level, and AI models predict what that sound is and translate it into a character in text, forming a string of characters used to build words, sentences and paragraphs; the most difficult part is converting a sound into a character, because no two people sound exactly alike and the same word can vary by speaker, origin, whether English is their first language, speaking speed and sentence position, and even the same person pronounces words differently at different times, such as when tired or speaking quickly. She says the technology has improved dramatically over the past few years partly because researchers continue developing better algorithms and because today's systems learn from vastly larger collections of speech than ever before, with users giving companies their audio to refine models whenever they use a speech-to-text system.