Google Launches Gemini 3.1 Flash TTS with Enhanced Emotional Expression and Multi-Speaker Capabilities

Gate News message, April 17 — Google unveiled Gemini 3.1 Flash TTS, an advanced text-to-speech model with enhanced emotional expression and control features, on April 15. The new model will be rolled out progressively through developer APIs, enterprise Vertex AI, and collaboration tools.

The model’s core capabilities include natural language-based audio tags for fine-tuning speed, intonation, and emotion, plus a “Director Mode” for specifying scenes and character roles to generate more nuanced voice outputs. A multi-speaker feature enables simultaneous dialogue generation, allowing more natural conversation flows suitable for podcasts, audio content, and AI assistants. The model supports over 70 languages and dialects, reflecting regional accents and expressions for localized voice experiences globally.

Google emphasized performance and cost efficiency, achieving high scores on blind human evaluation benchmarks while reducing computational costs through its Flash architecture—designed for large-scale enterprise adoption. Generated audio includes SynthID watermarking to identify AI-generated content and combat misinformation.

The move reflects intensifying competition in voice interfaces. OpenAI is combining real-time voice features with conversational AI for human-like interactions, while Meta is expanding investments in AI characters with voice-based social experiences. Industry observers note that while high-level acting and creative work may remain human-driven for now, repetitive and large-scale production markets could see gradual AI adoption in dubbing, advertising, and audiobook sectors.