Google launches Gemini 3.8 text-to-speech generation models

Google has officially unveiled two new text-to-speech models within the Gemini family: Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS. The release marks a strategic shift in AI voice generation, moving away from static audio presets toward highly customizable, studio-grade creative environments. Designed for creators, developers, and enterprises, the models deliver granular control over pacing, dialect, emotional tone, and backchanneling, enabling line-by-line performance direction across gaming, audiobooks, podcasts, and interactive media. The Flash variant prioritizes deep creative direction and character design, allowing users to generate entirely new vocal identities from natural language prompts. The Flash-Lite edition targets high-volume dubbing and cost-efficient scaling while maintaining expressive nuance. Both models expand access to over 2,000 production-ready voices covering more than 100 languages, including regional variants such as Mexican Spanish, Quebec French, and Scots English. A new voice replication feature enables developers to recreate consistent vocal profiles from 30-second audio samples, supported by upcoming voice remixing tools that will allow users to adjust timbre, pitch, and accent through text prompts. Performance benchmarks underscore the models technical advancements. Gemini 3.8 Flash TTS secured the top overall position on Hume AI's Voice Design Benchmark with a score of 71.4, while also leading in accent modeling. Both variants claimed the first and second positions on Hume AI's Overall Quality Index. In blind human preference evaluations on the Voice Arena platform, the models ranked highest among competitors across key global languages, including Japanese, Brazilian Portuguese, Vietnamese, Modern Standard Arabic, Mexican Spanish, and Hindi. Early testing indicates substantial improvements in long-form content generation and dual-speaker screenplay control compared to the Gemini 3.1 Flash TTS predecessor. Google has emphasized strict safety and transparency protocols surrounding voice synthesis. The voice replication system requires verbal consent verification matching the reference speaker before generation. All audio outputs are embedded with an imperceptible SynthID watermark and carry C2PA credentials to ensure detectability and combat misinformation. Development access is now live via Google AI Studio, which functions as a voice design workspace integrating prompt-based identity creation, voice replication, and dual-speaker screenplay editing. Through the Gemini API, infrastructure partners including Agora, LiveKit, Pipecat, and Vercel are enabling streamlined deployment of conversational voice agents and real-time speech interfaces. Google has also announced integrations with industry players such as Figma, HeyGen, Linguana, Wondercraft, 99.co, and Ollang to accelerate global media localization, nuance-driven regional dubbing, and scalable voice interaction systems. The launch reinforces Google's expansion within the Gemini Audio ecosystem, complementing recent releases like Gemini 3.5 Live Translate, 3.5 Transcribe, and 3.8 Live Extended Thinking.