Google is really pushing the boundaries with its speech models, releasing another Gemini 2.5 Flash Native Audio model last night.
It supports real-time speech translation, continuous listening, and bidirectional conversational speech-to-speech streaming translation, preserving the speaker's intonation, rhythm, and pitch.
It covers 70+ languages and 2000 language pairs.
Google Translate is available, initially on Android in the US, Mexico, and India.
twitter.com/op7418/status/1999...