Good question. If you look at Sondheim's work, he is the first to say he writes for actors who sing, not singers. Best example is Lansbury — not a trained singer, occasionally dubbed in her MGM films (i.e. THE HARVEY GIRLS) but allowed to sing for herself in less presentational, more British-centric pieces, like THE PICTURE OF DORIAN GRAY or TILL THE CLOUDS ROLL BY. On stage, she emerged as the prototype of the actor who sings. "The Worst Pies in London" could trip up the best of singers, but not an actor. So, barring an occasional "Green Finch and Linnet Bird," Sondheim usually writes for actors. INTO THE WOODS was written largely for actors who sing, but, based on his knowledge that Bernadette Peters was cast, his writing for the Witch is different. He knew he had a singer to work with, and the result is that her music is more challenging, more rangy, especially the bridge in "Stay With Me" and all of "Last Midnight." The reason they would have needed to dub those sections in the film version is because their name-above-the-title Witch was now an actor who sings. In Kendrick, they had a singing actress as Cinderella, and Pine is off the table because Marshall directed "Agony" broadly, almost for laughs, in the movie. All other characters in the film are actors who sing.
It was clearly Rob Marshall's intent that all of the actors sing for themselves in this film; Streep's dubbing likely became a necessity after the fact, especially given the pyrotechnic staging of "Last Midnight." If her vocal performance did not adequately merge with the imagery in rushes, something would have to be done. Once they had Murphy in the studio, they probably took inventory on all of Streep's vocals and decided to add the bridge of "Stay With Me" to her chores as a safeguard. And if you recall, it was that clip — the "Stay With Me" bridge — that was most often used by Disney during the film's hype period. It was the first clip released; you saw and heard it everywhere, presumably so that, in conjunction with the unveiling of Streep's distorting make-up, audiences would immediately accept the voice they were hearing as hers.
Updated On: 9/23/16 at 10:12 AM