Gemini 3.8 TTS Adds Custom Voices and 100+ Languages

Core News and Key Facts

The next big change in AI voice technology may be less about making a machine sound human and more about giving creators precise control over how that voice performs. Google has introduced Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, two new text-to-speech models designed to move beyond fixed voice presets. The models can generate original voices from natural-language descriptions and let users direct delivery, emotion, pacing and conversational cues.

Gemini 3.8 Flash TTS is available through the Gemini API and Google AI Studio for developers, while it is also rolling out to Gemini Notebook. Flash-Lite TTS is available through the same developer platforms and is coming to Google Vids. Enterprise API availability through Gemini Enterprise is planned for later.

Context and Official Statements

Google says Flash TTS is designed for detailed creative direction, including character voices, interactive media, podcasts and audiobooks. Flash-Lite TTS instead targets high-volume applications such as dubbing, audio creation and expressive voice agents. Both can control individual lines of dialogue and stage two-speaker conversations from a single script. Long-form generation is also designed to maintain voice consistency over extended audio.

One of the most notable features is voice replication. Google says an authorised voice can be recreated from a 30-second sample, but the system requires a consent recording from the voice owner. Generated audio is also protected with SynthID watermarking. Google says the models support more than 100 languages and dialects and provide access to more than 2,000 production-ready voices.

For Indian users, the language coverage is particularly relevant because Google specifically includes Hindi among the languages represented in its reported blind human preference evaluations.

Krihaa Analysis

The real shift here is from text-to-speech as a utility to AI voice as a production tool. Earlier TTS systems largely asked users to choose a voice and generate audio. Gemini 3.8 Flash TTS instead lets creators describe a vocal identity, direct individual lines and maintain that identity across longer projects. That changes the workflow for podcasts, audiobooks, games and video localisation.

For India, the dubbing angle could be particularly significant. A creator could theoretically develop a consistent character voice across multiple languages while controlling accent, pacing and emotional delivery. That does not eliminate professional voice artists, because performance direction, cultural adaptation and rights management remain important, but it could reduce the time and cost involved in producing multiple language versions.

The safety layer is equally important. Voice replication can create obvious impersonation risks, so Google requiring consent verification and embedding SynthID into generated audio is an attempt to make synthetic voices more traceable.

There is also a clear distinction between what is available now and what remains limited. Developers can already experiment with the new TTS models, but Google notes that voice replication through AI Studio is unavailable in India, among several other regions.

For Indian creators, therefore, the broader TTS capabilities are immediately relevant, while the headline-grabbing voice-cloning feature is still geographically restricted.