Research context and background

Google has launched two new text-to-speech models called Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS. These tools were released on September 23, 2026, as part of the Gemini Audio family. They represent a major step forward in how computers turn written text into spoken words. Instead of using simple, fixed voices, these models act like a creative studio. They allow users to have deep control over how a voice sounds, how it acts, and the emotions it conveys. This shift moves audio generation from static presets to a dynamic experience where every detail can be customized.

Benefits

The new Gemini 3.8 models offer several powerful advantages for creators and developers. First, they provide maximum voice fidelity and acting nuance. This means the generated audio sounds very natural and can capture complex emotions. The flagship model, Gemini 3.8 Flash TTS, supports over 130 languages, including many regional accents and minority dialects. It also maintains a consistent voice identity over long periods, which is crucial for projects like audiobooks or podcasts that last many minutes.

The second model, Gemini 3.8 Flash-Lite TTS, is built for speed and scale. It is optimized for high-volume production and low latency, making it perfect for tasks that require generating large amounts of audio quickly. Both models support a vast library of over 2,000 production-ready voices. Users can also create entirely new voices from scratch using simple text prompts. For example, a user could ask for a high-energy DJ voice from Melbourne or a monotone robot voice. The system also includes strict safety features, such as voice replication consent verification and audio watermarking to ensure content transparency.

Use Cases

These models are designed for a wide range of applications. Studio-grade audiobooks and complex multi-speaker dialogues are ideal uses for the flagship Flash TTS model. It excels at immersive gaming experiences and handling difficult pronunciation tasks. The ability to use stage directions and inline vocal events allows creators to add sounds like laughs, sighs, or gasps directly into the script for precise comedic timing.

The Flash-Lite TTS model is best suited for high-volume dubbing and everyday read-aloud features. It is also effective for building expressive voice agents that need to respond quickly. Both models support native multi-speaker staging, meaning a single script can drive a natural conversation with distinct voices taking turns. Long-form storytelling and podcasts that run for hours will benefit from the models ability to maintain high quality without voice drift.

Pricing

Pricing details for Gemini 3.8 Flash TTS and Flash-Lite TTS are not available in the provided information. The article mentions that the Flash-Lite model is designed for cost efficiency, but it does not list specific prices or subscription tiers.

Vibes

Public reception and specific user reviews are not included in the provided article. However, the models have achieved notable benchmark results. The flagship Gemini 3.8 Flash TTS ranked number one overall on Hume AI Voice Design Benchmark with a score of 71.4. It also took the top spot on Hume AI Overall Quality Index. The Flash-Lite model ranked second on the Overall Quality Index. These results indicate strong performance and high quality from the technology community.

Additional Information

Google is partnering with several companies to integrate these models into their products. Partners include Figma, HeyGen, Linguana, Wondercraft, 99.co, and Ollang. These collaborations aim to improve global dubbing, localization, and conversational voice agents. The models are available to developers immediately through Google AI Studio and the Gemini API. Enterprise users can access them via Google Cloud Vertex AI and Gemini Enterprise. The general public will gain access through the Gemini app. Voice replication features are currently restricted in Illinois, Texas, the EEA, the UK, Switzerland, and India due to regional regulations.