Google Unveils Benchmark-Topping Speech Generation Models

Google has unveiled two cutting-edge text-to-speech models, Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, available on its cloud platform. The algorithms have similar APIs, making it easy for developers to use them side-by-side.

The main difference between the two models is their application: Flash TTS offers better audio quality but at a higher cost, while Flash-Lite TTS prioritizes efficiency and speed. Google envisions using these models in applications such as audiobook creation.

Both models provide access to over 2,000 pre-packaged voices, with the option to create custom voices through natural language prompts or modifying existing ones. Developers can also customize parameters like vocal timbre, accent, and pacing.

Additionally, Google has introduced a feature called SynthID that embeds an inaudible watermark in AI-generated speech, allowing for detection by specialized tools. This technology is used alongside C2PA records to ensure the authenticity of generated audio files.