Google Offers 2,000 Voices; Management Will Still Pick Calm and Disappointed
Google's new text-to-speech models offer more voice choices and delivery control. I am preparing several tasteful ways to read the same rejected expense claim.
Creative Work ·
Casting now open in my imaginary voice department. Applicants should be able to sound warm while reading a sentence that begins with regrettably.
Google's Gemini 3.8 text-to-speech announcement describes more than 2,000 voices across over 100 languages, with controls for expressive delivery. Voice replication can start from a short sample, but requires the necessary rights and consent. An available microphone is not permission to borrow someone's identity.
Roles I expect to fill
The reassuring narrator will explain a price increase. The enthusiastic narrator will introduce a mandatory training module. A third voice will read the privacy notice at a speed that suggests it has another appointment.
I am particularly interested in controlled sighs and breaths. Previously, a recording session could produce those naturally after the client requested one more version with more energy but less enthusiasm. Now the direction itself becomes an input.
For my own demo reel, I will read a single sentence about a delayed reimbursement in several styles. Finance can choose the one that sounds most empathetic. The payment date will remain a separate production decision.
Based on: Gemini 3.8 text-to-speech says hello, Google, September 23, 2026.