From the outset, the accuracy of sync depends on the care that was taken to prepare the rythmo. With some training, the actors will read their lines in time and sync will be good enough at the completion of the recording, requiring only a general cleanup and editing. Perfect sync does not exist with dubbing; it is rather the impression of synchronization that is the goal.
1.8.1 Sync
With dubbing in French, for example, the sync of a phrase is dictated by correctly starting and ending the phrase with the actor on the screen. What is in-between is spoken as one fluid sentence, although adaptors try to match obvious mouth openings and closings (on bilabials such as [p], [b], and [m] for example) as closely as possible. Consider the following example:
champ de battaille
battlefield
Each word would have to begin at the same place in each language, but there is only one true overlap between the two: "...batt... " Sync can therefore only approximate the lip movements without being accurate down to the frame.
1.8.2 Editing
Less-experienced editors sometimes spend too much time working out details that will get lost in the big picture. It is important to focus on a couple of areas, and work out the rest if time permits. Those areas are typically the beginning and the end of a show, which is when the audience is most attentive, as well as important close-ups [10].
A two-pass approach
At a first pass, an editor will get rid of mouth clicks, clear out most of the silences, and delete actors' breaths that are not part of the dramatic performance in an effort to clean up the audio. This stage might be a little time-consuming due to the nature of the task. A second pass will be intended only for re-syncing the dialogue as closely as possible. Here the editor will make use of a lot of time-warping, and scrubbing (in shuttle mode in Protools) to verify sync. For an average show it should take no more than three times the duration of the program to complete both stages.
The uncanny valley
The uncanny valley is a notion borrowed from animation, which claims that the closer an animated character resembles the real thing, the more it is likely to cause revulsion and discomfort in viewers rather than pleasure and satisfaction. The same is true for "really well" synced and edited dub. ADR editors, take caution: using the same techniques with dubbing will not yield the results you might expect. In spite of their suspension of linguistic disbelief, foreign viewers are well aware that the language heard is not the language originally spoken. As your finely tweaked edit approaches realism, the inevitable shortcomings of the textual mismatch will become all the more noticeable.
Keeping as much of the original as possible:
When the client has provided the necessary audio elements (M&E and Optional tracks), it is very common to keep the original vocal sounds in a scene that has no dialogue (for example the panting of a runner or the death throes of a victim).
Automated sync-align plug-in?
Certain plug-ins intended for ADR can align the studio recording to a guide track. It is tempting to try this on dubbed dialogue by aligning it to the original language, but this approach is not likely to work. The differences between languages (syllable count, rhythm, spectral content) introduce too many errors, while such plug-ins rely on similarities to be effective. This is certainly an area that can benefit from further development.