Artificial Intelligence
Google launches Gemini 3.8 text-to-speech models
Google introduced Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, adding custom voice design, voice replication and line-by-line direction, and said they were rolling out across its products and developer tools.
- Google introduced Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS.
- Google said Flash TTS was for custom voice creation and detailed performance control, while Flash-Lite was aimed at high-volume dubbing and voice agents.
- Google said Flash TTS scored 71.4 and ranked first on Hume AI’s Voice Design Benchmark, and that the two models ranked first and second on Hume AI’s Overall Quality Index.
- Google said generated audio used SynthID watermarks and that voice replication required a verbal consent recording from the voice owner.
- Google said the models were available from that day in the Gemini API and Google AI Studio, with other Google products to follow.
Google introduced two text-to-speech models in the Gemini family: Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS. It said Flash TTS was aimed at custom voice work and detailed direction, while Flash-Lite was aimed at higher-volume, lower-cost speech generation.
The company said Flash TTS could create voices from natural-language prompts in more than 100 languages and dialects. It said users could also draw from more than 2,000 production-ready voices or recreate a voice from a 30-second sample.
Google said users could save custom voices for later use and adjust timbre, pitch, pace and accent. It also said both models supported long audio runs and two-speaker scenes, including non-verbal cues and back-and-forth sounds.
Google said Flash TTS scored 71.4 and ranked first on Hume AI’s Voice Design Benchmark, and that Flash TTS and Flash-Lite ranked first and second on Hume AI’s Overall Quality Index. It said generated clips were watermarked with SynthID, and that voice replication required verbal consent from the person whose voice was being copied.
Google said Flash TTS would be available in Gemini Notebook, while Flash-Lite would be available in Google Vids, and that enterprise access would come through Gemini Enterprise. It also said developers could use the models in Google AI Studio and the Gemini API from that day.
Named in this story
Companies
- announced the models and described their rollout
- Hume AI
- the company whose benchmarks Google cited
Products and systems
- Gemini 3.8 Flash TTS
- the new model for custom voice design and performance control
- Gemini 3.8 Flash-Lite TTS
- the new model for high-volume speech generation
- Gemini API
- the developer interface where the models were made available
- Google AI Studio
- the developer workspace where the models were made available
- Gemini Notebook
- one of the Google products set to use the models
- Google Vids
- another Google product set to use the models
- Gemini Enterprise
- the enterprise route Google said would get the models soon
- SynthID
- the watermarking system Google said it applied to generated audio
How the source tells it
The piece read like a product launch post, with sales language, benchmark bragging and a safety section used to reassure.
- vendor boosterism self-promotional first-person launch language around its own models and products
- hype expansion claims and superlatives used to present the features as transformative
- competitive triumph rank claims and benchmark scores used to frame the models as winners
- reassurance safety, consent and transparency controls used to soften the release