Gemini 3.8 TTS: Google Brings Custom Voices to AI Audio

Core News and Key Facts

AI voice generation is moving beyond choosing a preset voice, and Google’s latest Gemini release is making the voice itself part of the creative process. Google has introduced Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, two text-to-speech models designed for expressive audio generation, custom voice creation and large-scale production. The models became generally available through the Gemini API and Google AI Studio in September 2026. Flash TTS is also rolling out to Gemini Notebook, while Flash-Lite is being added to Google Vids.

The flagship Flash TTS model is aimed at high-quality creative work such as audiobooks, podcasts, studio narration and complex dialogue. Flash-Lite is positioned as the faster and more cost-efficient option for high-volume generation, read-aloud features and conversational voice agents. Google says Flash TTS supports more than 130 languages, while Flash-Lite supports 101, including Indian languages such as Telugu, Hindi, Tamil, Kannada and Malayalam.

Context and Official Statements

One of the biggest changes is the ability to design a voice using natural-language instructions. Developers can describe characteristics such as accent, vocal style and delivery rather than simply selecting from a fixed list. Google also supports voice replication from authorised samples, with consent verification intended to ensure that a voice owner has approved the replication. Generated audio carries SynthID watermarking.

The models can also stage two-speaker conversations from a single script while maintaining separate voices. Individual lines can receive instructions for tone, pacing and emotional delivery, while tags can introduce sounds such as laughs, sighs and gasps. Google says Flash TTS is designed to maintain voice identity and consistency across longer recordings. However, an important India-specific limitation remains: Google states that voice replication through AI Studio is not currently available in India.

Krihaa Analysis

The more important development here is not simply that Gemini can produce a more natural voice. It is that Google is turning TTS into something closer to an audio production workflow. For Telugu creators, this could eventually reduce the gap between writing a script and producing a finished regional-language audio product. News explainers, educational content, podcasts, dubbing and accessibility features can potentially be generated without recording every line manually.

The distinction between Flash and Flash-Lite also reveals where Google expects demand to grow. Flash is being positioned for quality-sensitive creative production, while Lite targets scale. That matters because the next wave of AI audio is likely to involve thousands of short voice interactions rather than only a few polished demonstrations.

At the same time, voice replication introduces a much bigger question than audio quality: ownership. Consent verification and SynthID are important safeguards, but they do not eliminate the need for creators and platforms to establish clear permission before reproducing identifiable voices. For Indian users, the current unavailability of AI Studio voice replication also means the headline feature is not yet fully usable locally.