A position is shown, then hidden. A series of moves are listed that you must visualize mentally. In the resulting position, find the winning move. Difficulty depends on the number of moves and puzzle complexity.
Pro tip: Visualize each move one at a time, updating your mental board before moving to the next.
A position is shown, then hidden. A sequence of moves is listed that you play out in your head, and in the position that results you have to find the winning move. It stacks two skills that most training keeps separate: holding an updated board, and calculating inside it. Either alone is manageable. Doing both at once is what actually happens in a blindfold game, and it is why players who ace static recall still fall apart the moment a game starts moving.
The discipline that makes it work is refusing to look for the tactic while you are still updating. Play the sequence to the end, take a beat to let the final position settle, and only then start searching. Players who try to calculate as they go end up solving a puzzle in a position that no longer exists. Once the picture is stable, run the normal checks in order: forcing moves first, checks and captures before quiet moves, and be especially careful about pieces that moved during the sequence, since the tactic almost always involves one of them. Practise the update in isolation with position reconstruction and the calculation in isolation with blindfold tactical puzzles. When both feel routine, you are ready for a full blindfold game.