I tried subtracting vector3 position of the card (since it moves with a bit of delay) from vector3 position of the cursor, multiplying it a bit and applying x and y of the multiplied vector difference to x and y rotation of the card, but i'm afraid it's not good.
Maybe i can try getting current and last frame pointer position, subtract, multiply and apply to rotation, but that's a similar approach and i'm not sure it will work.
Or maybe a separate FSM with clamped object rotation by mouse.