- Google has officially announced its latest text-to-speech AI models, “Gemini 3.8 Flash TTS” and “Gemini 3.8 Flash-Lite TTS.”
- “Gemini 3.8 Flash TTS” is a voice generation AI model designed for creative applications such as games, audiobooks, and podcasts, with a strong focus on character voice creation and detailed direction.
- “Gemini 3.8 Flash-Lite TTS” is designed with a focus on scale and cost-efficiency for large-scale dubbing, audio content production, and expressive voice agents. Despite its low cost, it allows control over voice tone, speaking pace, and subtle nuances.
Just after announcing “Gemini 3.8 Live” and “Gemini 3.8 Live Extended Thinking”, AI models specialized for voice conversation, Google officially announced its latest text-to-speech AI models, “Gemini 3.8 Flash TTS” and “Gemini 3.8 Flash-Lite TTS,” on Wednesday, September 23, 2026 local time.
Both “Gemini 3.8 Flash TTS” and “Gemini 3.8 Flash-Lite TTS” are their most expressive voice generation models to date. They go beyond simply reading text using pre-set voices, enabling studio-like use cases such as designing entirely new voices or giving acting instructions for specific lines. Rollout has already begun through the “Gemini API,” “Google AI Studio,” “Gemini Notebook,” “Google Vids,” and more.
Additionally, built-in safeguards to prevent misuse are included: all audio generated by both the “Gemini 3.8 Flash TTS” and “Gemini 3.8 Flash-Lite TTS” models will feature “SynthID,” an imperceptible digital watermark, and voice replication requires a verbal consent recording from the actual voice owner.
“Gemini 3.8 Flash TTS”
“Gemini 3.8 Flash TTS” is a voice generation AI model tailored for creative applications like games, audiobooks, and podcasts, focusing on character voice design and detailed direction.
By simply specifying roles, accents, and voice qualities in natural language, you can create original voices from scratch supporting over 100 languages and dialects. You can also choose from more than 2,000 ready-made voices, including regional variations like Mexican Spanish, Quebec French, and Scottish English. Furthermore, if it is your own voice or one you have the rights to use, you can recreate a voice from just a 30-second audio sample and save it to maintain consistent voicing across ongoing projects.
A “Voice Remix” feature is also planned for future release, allowing users to fine-tune the tone, pitch, speaking speed, and accent of ready-made voices using prompts.
“Gemini 3.8 Flash TTS” also allows you to add stage direction-like acting instructions for specific lines to precisely control delivery, generate dialogue between two speakers from a single script, and insert non-verbal expressions such as laughter, sighs, or acknowledgments. Google states that the model maintains consistent voice quality and pacing even across hours of long-form audio.
These acting instructions, long-form generation, and dual-speaker dialogues are also available in “Gemini 3.8 Flash-Lite TTS,” detailed below.
The rollout of “Gemini 3.8 Flash TTS” has begun for developers via the “Gemini API” and “Google AI Studio,” and for general users via “Gemini Notebook.” It will also be made available to enterprise customers via API through “Gemini Enterprise” in the future. Note that the voice replication feature in “Google AI Studio” is not available in Illinois, Texas, the EEA, the UK, Switzerland, and India.
“Gemini 3.8 Flash-Lite TTS”
“Gemini 3.8 Flash-Lite TTS” is a voice generation AI model designed with scale and cost efficiency in mind for large-scale dubbing, audio content production, and expressive voice agents, offering control over voice tone, speaking pace, and subtle nuances at a low cost.
The rollout of “Gemini 3.8 Flash-Lite TTS” has begun for developers via the “Gemini API” and “Google AI Studio,” and for general users via “Google Vids,” with future availability for enterprise customers via API through “Gemini Enterprise” as well.
Source: Google (https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-text-to-speech/)