Google has released two Gemini voice models that support custom voices and multi-role dialogue.

Google launched two new text-to-speech (TTS) models on Wednesday: Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, further expanding the application of Gemini in the field of AI speech generation.

Flash TTS focuses more on sound performance and complex creative scenarios, while Flash-Lite TTS is optimized for large-scale, low-cost speech generation, with applications including content dubbing, audio production, and real-time speech proxy.

Google developer documentation shows that Gemini 3.8 Flash TTS supports 130 languages, while Flash-Lite TTS supports 101 languages. It also allows users to adjust voice, speech rate, emotion, and dialogue style based on text commands.

Flash TTS focuses on customized sound, ranking first in overall sound design metrics.

According to Google, Gemini 3.8 Flash TTS is primarily designed for creative applications such as audiobooks, podcasts, game characters, and interactive media. Users can design voices from scratch using natural language prompts, including character settings, accents, and voice characteristics.

The model also supports copying sound from approximately 30 seconds of audio samples. Google states that obtaining the owner's consent and passing an appropriate consent verification mechanism is required before copying the sound. The generated audio will then be embedded with an imperceptible SynthID watermark to help identify the AI-generated speech content.

The Gemini 3.8 Flash TTS scored 71.4 points in Hume AI's Voice Design Benchmark, ranking first in the overall sound design index; the two models ranked first and second respectively in Hume AI's Overall Quality Index.

Currently, Google offers more than 2,000 ready-to-use voices, covering regional language variations such as Mexican Spanish, Quebec French, and Scottish English.

Both models enhance multi-role dialogue.

In addition to voice customization, both models support individual control over each line of dialogue, allowing users to adjust tone, rhythm, emotion, and pauses through script commands.

Google says the new model also supports two-person conversations and can maintain the distinction between the voices of different characters within the same script. Furthermore, the model can generate nonverbal sounds such as laughter and sighs, as well as conversational responses like "um" and "yes," making the generated speech more realistic.

Google says both models are now available to developers through the Gemini API and Google AI Studio.

Flash TTS can also be used with Gemini Notebook, while Flash-Lite TTS is already integrated with Google Vids. Support for the Gemini Enterprise API will be released later.

From a product positioning perspective, this Google update is not simply about improving the naturalness of text-to-speech, but rather an attempt to further enhance users' control over the "design" and "performance" of AI voices, expanding voice generation from fixed timbre selection to customizable characters, scenarios, and multi-character dialogues.

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.