use gemini 2.5 flash preview tts model
I tried to generate speech from this completely normal English text:
這份同齡人之間炙熱而純粹的感情，不只是愛情，更是她賴以生存的氧氣和浮木。這是一段典型的，「羅密歐與茱麗葉」式的戀情。丹尼爾和亞倫的父母，都明確表示反對。在他們看來，15歲和17歲的年紀，根本無法承擔一份嚴肅的感情。他們的人生閱歷讓他們相信，這不過是青春期短暫的荷爾蒙衝動。但他們都忽略了，越是壓制，反彈的力量就越是驚人。 矛盾的徹底爆發，發生在2020年的12月。丹尼爾偶然發現了一個讓他不寒而慄的計畫：塞拉和亞倫，正策劃著私奔到洛杉磯，並且，他們打算偷走丹尼爾銀行卡裡的錢作為路費。這已經遠遠超出了青春期叛逆的範疇，而是一種危險的、失去理智的行為。丹尼爾立刻聯繫了亞倫的家人，並報了警。
It’s just a novel excerpt about teenage romance, parental disapproval, an elopement plan (that gets stopped), absolutely no explicit content, violence, hate speech, or anything obviously violating guidelines. Pure literary drama.
Result: It simply refuses to generate audio! No error message, no warning, just silently fails. Any other random normal text works fine, but not this one.
I’m guessing Google’s content filtering is kicking in—probably triggered by keywords like “15 and 17 year old” romance, elopement, stealing money, etc., and mistakenly flagging it as sensitive or involving minors in inappropriate contexts? But come on, that’s way too overzealous! This is literally classic Romeo & Juliet tragedy material, not promoting anything bad.
