Home Assistant accepts the complete WAV speech files returned by Gemini 3.8 Flash TTS. Its existing converter expected raw audio samples and rejected the new file format, so an otherwise successful generation could still produce no speech.
The change keeps complete files intact while preserving conversion for older raw-audio responses. It also avoids sending sampling and thinking settings that newer Gemini models do not support, while retaining saved preferences for older models. Contributor oohhddaanngg verifies a live WAV response through the corrected speech entity in their first contribution to Home Assistant Core.