Google has rolled out substantial updates to its Gemini 2.5 Text-to-Speech (TTS) models, specifically the Flash and Pro versions. The enhancements focus on expressivity and control, crucial for developers needing high-quality voice synthesis. Key improvements include richer tonal variations, precise pacing that adapts to context, and more consistent character voices in multi-speaker settings. These updates are designed to improve applications ranging from audiobooks to e-learning modules and marketing content, with the models expected to replace earlier versions released in May. The Gemini 2.5 models now allow for highly specific tone requests, enabling developers to create characters with distinct personalities. Users can ask for tones like "cheerful" or "somber," and the model will adjust accordingly. Pacing adjustments are also more sophisticated, enhancing the naturalness of speech by slowing for dramatic moments or speeding up during action sequences. The multiple-speaker capabilities have been refined to maintain unique character identities and smooth transitions in dialogues, critical for podcasts and narratives. The multilingual features support consistent tone and pitch across 24 different languages. Companies like Wondercraft and Toonsutra are already leveraging these advancements. Wondercraft employs the models for creating realistic multi-speaker conversations and precise control over speech elements, while Toonsutra uses them for cinematic voiceovers in multiple languages. Developers can access the new Gemini 2.5 models through the API in Google AI Studio, with resources available to help them get started quickly.