Google introduced Gemini 3.8 Flash TTS and Flash-Lite TTS, speech-generation models for designing original voices, reproducing authorized vocal profiles from a 30-second sample, and directing performances with natural-language instructions. Google says replicated voices require consent verification, while generated audio includes SynthID watermarking and C2PA provenance credentials. The models are available through Google AI Studio and the Gemini API, with integrations in Gemini Enterprise, Gemini Notebook and Google Vids. Google reports a 71.4 overall score on Hume AI's RW-Voice-EQ benchmark, but production teams still need to test language, accent, consistency and latency for their own workloads. The safeguards reduce but do not eliminate impersonation risk: watermark detection and provenance metadata depend on downstream platforms preserving and exposing them.