Lipsync is a video-processing operation: you have a recording and you have a new audio track, and the model reshapes the speaker's mouth so that it matches the new words. Without it a dubbed video gives itself away immediately — the viewer may not be able to say what is wrong, but can see that something is.
The most important practical point is that this is two separate services, not one. The audio side — translation and speech generated in the speaker's own voice — and the picture side — reworking the mouth movement — are different tools, and the largest voice provider states in its own help pages that its dubbing does not move the mouth.
Out of that comes the most common expensive mistake: someone buys a dubbing tool believing the picture will be taken care of as well, and then finds out that the video needs a separate tool and a separate budget. Run the test on a short talking-head clip before you subscribe to anything.
The full workflow — what to use for what, and in what order — is the subject of Lipsync and Dubbing: Your Video in Another Language, a post of its own. Directly related terms: voice cloning, AI voiceover and transcription.