Uh, no? This is not at all an absurd requirement? Screen readers literally do this all the time, with voices that are the classic way of making a speech synthesizer, no AI required. ESpeak is an example, or MS OneCore. The NVDA screen reader has an option for automatic language switching as does pretty much every other modern screen reader in existence. And absolutely none of these use AI models to do that switching, either.