The State of Modern AI Text to Speech Systems for Screen Reader Users
Summary
The article examines the state of AI-based text-to-speech for screen reader users, arguing that recent neural TTS developments have largely sidelined accessibility needs. It highlights persistent issues such as dependency bloat, accuracy problems (skipped words, mispronunciations), speed limitations for screen readers, and lack of user-adjustable voice parameters, with specific critique of Eloquence, Espeak-NG, Kitten TTS, and Supertonic, and discusses possible directions for improvement and community efforts.