Voice Design previews now use Gemini 3.1 Flash TTS (preview) by default. It is the engine that generates the preview when you design a voice.

The switch is for faster, higher-quality previews, so the sample you hear while designing comes back quicker and sounds better than it did on the previous default.