Gemini 2.5 Pro Preview TTS

Overview. Gemini 2.5 Pro Preview TTS is a proprietary speech synthesis model developed by Google for generating high-fidelity audio from text prompts. Built as a specialized audio variant of the Gemini 2.5 family, it accepts up to 8,192 text input tokens and synthesizes up to 16,384 tokens of linear PCM or WAV audio. Developers deploy this model for direct, turn-based spoken audio generation within Google-backed inference pipelines.

Key strengths. The model focuses entirely on text-to-speech rendering, supporting extended spoken output up to its 16,384 audio token maximum. Its dedicated audio synthesis pipeline converts raw text into clear vocal output without requiring external audio post-processing. Output is delivered directly in standard audio formats, streamlining ingestion into downstream playback and media systems.

When to use it. Use this model for turn-based voice synthesis workloads such as spoken content generation, voiceover narration, and offline automated messaging. It is appropriate for interactive pipelines that submit prepared text blocks and receive complete audio streams. For long-term production deployments, plan migration paths to Gemini 3.8 Flash TTS ahead of Google's scheduled November 2026 retirement date.

Limitations. The model is strictly limited to text input and audio output, lacking capabilities for text generation, function calling, structured outputs, or context caching. It does not support real-time bidirectional audio streaming through the Live API, and Google has slated the preview endpoint for shutdown on November 17, 2026.

Sovereignty note. Hosted on Google Cloud infrastructure under United States jurisdiction, this model is subject to the US CLOUD Act and does not offer Sovereign European operational isolation.