Narrative Voice
Consistency is not the goal — intentionality is
Consistency Is Not the Goal
It would be easy to build a tool that flags every place your voice changes and calls each one a problem. That tool would be wrong about most of them. A manuscript whose voice deliberately ranges — dark violence set against comedic warmth, a limited narration opening briefly into omniscience for a single interlude — would light up end to end, and the three genuine breaks would be lost among a hundred intended ones.
So the standard is not uniformity. The standard is that every shift is one you chose. A deliberate change of register at a turning point is craft. The same change appearing because you drafted that chapter in a different month is drift, and readers register it as a wobble in the storytelling even when they cannot say what moved.
Judging which is which is your call, not the model's — it depends on intent, which is not in the text. That is also why Bookshaper does not score it: a range-not-fault signal cannot support a verdict, so there is no voice reading in the app to argue with. What follows is how to do the judging yourself.
Fixing Drift
Voice repairs are line-level work, and they go faster with a reference than with a rule. Pick two or three pages of your own prose that sound most like the book you meant to write, and keep them open while you revise the chapter that drifted. You are matching a sound, not applying a standard.
Work in one pass per dimension rather than trying to fix everything at once. A pass for sentence rhythm, then a pass for register, then a pass for narrative distance, will get you further than a general attempt to make a chapter sound right.