Text to speech: turn up to 2,000 characters into natural-sounding speech (MeloTTS), returned as base64 MP3. Languages: English, Spanish, French, Chinese, Japanese, Korean.