Text-to-speech (TTS) is the AI technology that converts written text into natural-sounding spoken audio, and in e-learning production it is the voice layer of AI avatar workflows and narrated slide videos. Modern neural TTS systems — such as those behind ElevenLabs, Azure Neural Voice, and Amazon Polly Neural — produce speech that is often indistinguishable from a human reader on a first listen, with controllable prosody, pace, and emotional tone. The workflow for a course update is compact: a subject-matter expert edits the script, the TTS engine renders the new audio in seconds, and the AI avatar or screen-recording tool syncs the video to the new audio track. This makes iterative course maintenance practical at a frequency that studio recording does not allow. TTS and ASR (Automatic Speech Recognition) are complementary: ASR converts speech to text (transcription direction), while TTS converts text to speech (synthesis direction); together they enable translation and re-narration workflows for multilingual localization. Consent and voice-cloning ethics are the key governance concern: cloning a specific person's voice for course narration requires explicit rights, and synthetic voices should be disclosed to learners. Audio quality varies with the underlying model and with prosodic complexity — technical jargon, acronyms, and non-English names often need manual pronunciation adjustments.
