Voice AI (TTS · STT)

Voice AI works in two directions. Turning speech into text is STT (Speech-to-Text); turning text into speech is TTS (Text-to-Speech). When a meeting recording automatically becomes minutes, that's STT. When navigation software reads directions aloud, that's TTS.

The recent leap forward is in sounding "human." A voice can now be cloned from just a few seconds of sample audio (voice cloning), complete with breathing, laughter, and hesitation. This is what makes chatbots and real-time voice conversations possible, and what has given rise to AI dubbing and AI audiobook industries.

The shadow side is just as clear. Voice-phishing scams that imitate a family member's voice, livelihood concerns among voice actors, controversies over recreating the voices of the deceased — voice AI stories tend to pair convenience with cases of misuse.

"New TTS clones a voice from a 3-second sample" — bragging about performance and warning about misuse tend to show up together in this kind of story.