Improving Gemini Text-to-Speech models for better control and capabilities

Today, we’re announcing significant enhancements to our Gemini 2.5 Flash and Gemini 2.5 Pro Text-to-Speech (TTS) preview models.

Key improvements include:

• Enhanced expressivity: Richer tone versatility and stricter adherence to style prompts.
• Precision pacing: Smarter context-aware speed adjustments and better instruction following.
• Seamless dialogue: Consistent character voices in multi-speaker scenarios.

These models will replace our TTS models released in May. You can start vibe coding apps in Google AI Studio and explore model capabilities today in the Playground.

New TTS Flash and Pro models

Many developers rely on text-to-speech for generating high-fidelity content that requires granular control over style, tone, pace, and accents—from long-form audiobooks to localized e-learning modules. Use cases like product tutorials, marketing videos or creator content also often require multiple voice interactions and reliable technical pronunciations.

To address these needs, we are launching updates to both Gemini 2.5 Flash TTS preview (optimized for low latency) and Gemini 2.5 Pro TTS preview (optimized for quality).

Enhanced style and tone versatility

Whether you are building a role-playing game character, a helpful virtual assistant, or a dramatic narrator, the voice needs to fit the role. Our Gemini TTS models are now far more expressive and align much more closely with specific instructions provided in your style prompt, dramatically improving role adherence. You can request a specific tone—from "cheerful and optimistic" to "somber and serious"—and the model will deliver a performance that feels authentic to that instruction.

Explore the Gemini 2.5 TTS model’s range of styles in action in our Synergy Intro demo app.

Context-aware pacing control

To create natural speaking patterns, pacing is a critical element: a joke needs timing, a complex explanation needs room to breathe, and an action sequence needs speed. We have refined the model's ability to adjust pacing based on the context of the message, allowing it to naturally slow down for emphasis or speed up for excitement. Furthermore, we've improved pacing control, meaning the model now follows your explicit pace-related instructions with much higher fidelity.

Listen to how the model adjusts its pacing according to the instructions:

"Style: You are a storyteller for a mystery novel. Start with a nervous tone that accelerates into excitement and relief"

Text: "I tried unlocking the door slowly. I fiddled around nervously... nothing... breathing deeply, I tried a second time. Click, I got in! I can't believe it, I actually got in!"

Improved multiple-speaker capabilities

For use cases like podcasts, simulated interviews, or multi-character narratives, creating realistic dialogue with distinct identities is key. We have refined the models to maintain consistent character voices and handle the "handoff" between speakers more naturally during back-and-forth exchanges.

We've also improved the multilingual capabilities of our model, so that it can preserve the unique tone, pitch, and style of each character throughout the conversation across all 24 supported languages.

Explore the model’s multi-speaker, multilingual capabilities in our Voices from History demo app.

What customers are saying

Partners are already seeing the impact of these improvements in production.

Wondercraft, an AI audio platform, uses Gemini TTS to build two of their most important features: Convo Mode, which lets anyone create life-like multi-speaker conversations with control over pacing and delivery, and Director Mode, which gives precise control over pronunciations, intonation and non-verbal cues, making edits feel effortless.

Toonsutra brings their stories to life with cinematic voiceovers for characters and promotional video ads, relying on Gemini TTS for its ability to handle diverse languages and character nuances.

Get started today

Gemini 2.5 Flash TTS and 2.5 Pro TTS models are available via the Gemini API in Google AI Studio. Read our developer docs, explore the prompting guide, or check out the Gemini API Cookbook to get started.