To explain it further requires some jargon. There is a difference between an object language - a language we use to make statements - and a metalanguage - a language we use to talk about statements in the object language.
Tarski's T-schema are perhaps easier to see as being non-trivial if we use a different language for the object language. For example,
"La neige est blanche" is true in French if and only if snow is white.
Note that the left part is in quotes and the right part is not. The quoted part is a sentence in the object language; the rest is a sentence in the metalanguage. The sentence as a whole is stating (in English) under what conditions "La neige est blanche" is true (in French).
But even if we use English in the object language, it is not trivial.
"Snow is white" is true in English if and only if snow is white.
This sentence is stating in the metalanguage under what conditons "snow is white" is true.
Snow is white
is a statement about snow.
"Snow is white" is true
is a statement about the truth value of the sentence "snow is white".
If this still seems rather trivial, the payoff is that a language with a truth predicate, i.e. a way of saying "X is true", is stronger than without. In fact, if a (first order) language is strong enough to express arithmetic, Tarski showed that truth is undefinable within that language. This can be stated formally by saying that no consistent theory can contain all instances of the scheme:
True(⌜φ⌝) ↔ φ
Where φ is some sentence in the language, ⌜φ⌝ is the quoted sentence, and ↔ is the material biconditional. Informally, this means that the concept of truth within a first-order language cannot be defined by a formula within that language.