The Dialogue Needs Repair. Will Re-recording Preserve the Performance You Pitched?
The Dialogue Needs Repair. Will Re-recording Preserve the Performance You Pitched?
Somewhere in the review notes there is a sentence like can't hear the line, we'll fix it in post. That sentence is doing two jobs, and only one of them is true right now. It is describing a problem: the words are hard to understand. It is also promising an outcome: replacing the words will leave the approved performance intact. Those are different claims. A clearer line is not automatically the same performance, and until someone has actually heard both against the picture, nobody knows which one they are getting.
So the useful move is not to argue for or against the booth. It is to name what the take is doing, ask the sound team what this particular recording can realistically retain, and then run three routes against the same picture: a repair of the original, a newly performed line, and an edit that reduces the need for the damaged passage. Each can solve intelligibility. Each can also change something the treatment already sold.
Identify what is damaged and what has to survive
Start by separating the words from everything the words are attached to. A performance carries information beyond its transcript: the breath a performer takes before a claim, a reply that steps on the end of another line, the hesitation that a partner answers, the strain in a voice that is doing something physical. Those are not decoration on top of the dialogue. They are frequently the reason the shot is in the cut.
Consider a constructed scene. Two performers carry a lightweight prop box across a set while acting effort. The first performer says "I've got it," catches a breath, and hesitates. The second performer's grip changes in answer to that hesitation. Nothing here has been recorded—this is a comparison sketch, not a listening result—but the dependency is the point. The words "I've got it" are maybe a third of the cue. The rest is a breath that reveals the strain, a pause that asks a question without words, and a hand that answers it. If you only hear the failure, you will treat all of that as the noise around a broken line. It is closer to the line's job.
So build an inventory before anyone touches the file. What is actually unintelligible—one word, a phrase, or the whole line? Is the failure in the recording, or in the performance? A word swallowed by a bad boom position is one problem; a word the performer chose to throw away is another, and the second one may be doing expressive work. Then list what must survive a change: the breath before the claim, the overlap with the partner, the length of the hesitation, the moment the grip changes.
That list is what you are protecting. It is also what makes the rest of the conversation testable instead of taste-based.
A caution worth stating plainly: perfect verbal clarity is one approval criterion, not the only one. Deliverables get reviewed for "can I hear it," which is easy to answer, and rarely for "does the relationship still read," which takes an afternoon. If the second question never gets asked, the first one wins by default.
Ask what the original can realistically retain
Before the booth, ask the sound team to assess a bounded repair using the actual recording. Bounded means specific: can these words be made intelligible without altering the breath that carries them? Where the breath sits underneath the word, the two may not be separable—fixing one may take the other with it. That is a question for someone who can hear the file, not something a treatment can assume.
Ask for two things back. First, which problems the repair leaves unresolved. Second, which expressive details the repair risks changing along the way. Then insist on an audible comparison against the picture, not a verdict. "Cleaned up" is a description of the process. It is not evidence that the performance survived the process.
There is a related option worth checking before any new recording: the session. A different take of the same line is still the same performer, the same day, the same body under the same box. If the material exists, an alternate is the closest thing to the original performance short of the original itself. It is also a different performance, with different timing, so it goes through the same comparison as everything else.
The practice is worth naming before it gets mistaken for a guarantee. The habit is to retain production recordings where possible, draw on alternate takes, and replace selected words or portions of a line when needed—while treating the captured performance and its surroundings as the thing worth protecting. The ambition is word-level, not line-level, and the recording is presumed valuable until proven otherwise.
Two limits matter. Practice is not your session, and a habit is not a result. It can tell you that selective replacement is a real discipline; it cannot tell you that your specific repair succeeded. Only the comparison can.
Compare a new performance and an editorial alternative
Now the fork. Route one is the repaired original. Route two is a newly recorded line. Route three is an edit that shortens or avoids the troubled passage.
For route two, the new line is not just words matched to a mouth. Play it against the selected picture and watch four things at once: the timing, the physical effort, any overlap with the other performer, and the partner's response. In the box sketch, ask whether the replacement "I've got it" still arrives mid-strain, or whether it lands as a clean statement with a body missing underneath it. A booth read can be perfectly intelligible and still change the apparent intention—reassurance becomes assertion, effort becomes ease.
The overlap is the trap most teams underestimate. If the first performer's line runs under or over the second performer's line, then the second performance is fixed, and the new first line has to fit a slot defined entirely by someone else's timing. You are not recording a line. You are recording one voice into a gap that another voice already shaped.
For route three, the edit solves intelligibility by removing the need for it. That is a real option, and sometimes the right one. It also has a specific cost: it can remove the cause of a reaction. Trim the hesitation in the box sketch and the grip change loses its trigger. Ask the direct question of the shot that follows—does the partner now read as responsive, as distrustful, or as simply unmotivated? Those are three different scenes, and the audience will pick one whether or not anyone intended it.
The three routes also carry different approval footprints. Repair stays inside sound. A new line stays inside sound but introduces a new performance that someone with creative authority should hear. The edit changes picture, which means it goes through whoever approved the cut. Knowing which door each route walks through is part of deciding between them.
Approve the consequence, not just the cleanest line
When the routes go to a decision, present them with the same relevant context: same picture, same passage, same playback conditions. A clean line played in isolation will always win an argument against a rougher one played in isolation. That is not a comparison. It is a demonstration of the thing you already decided.
Keep the three questions separate, because they are usually averaged into one. The first is intelligibility: can the words be understood. The second is performance continuity: does the voice still belong to the body and the moment it was recorded in. The third is causal logic: does the reaction that follows still have a visible cause. A route can win the first and lose the second and third. Averaging them into "sounds good" hides the trade you are actually making.
Then name who approves a changed creative reading—the person with authority over the take's intent, the same authority that signed off on the performance in the first place. A replaced voice is a new choice, even when the words are identical. It should not be approved by whoever happens to be in the room when the file gets dropped in, and it should not be approved by a description.
The sentence to avoid is the one that ends the conversation: something like "dialogue restored" or "indistinguishable from the original." Neither can be stated before the work is heard, and the second is a very high bar. A repaired recording and a re-recorded one can both be perfectly good and still be two different takes. That is not a failure. It is the decision.
What to write, and what to hear first
A treatment can promise a process. It cannot certify an outcome nobody has heard. The honest version names the work: repair the original where the performance depends on it, replace only what cannot be recovered, and treat any edit that removes the troubled passage as a change to the scene's logic—then verify all of it against the picture.
In the box sketch, that means a defined comparison. Does the repaired original still let the breath carry the words? Does the replacement line still arrive mid-strain, still fit the partner's timing, still earn the grip change? Does the shortened edit leave the partner reacting to nothing? Those are the questions. They are not results, because no scene, no damaged recording, and no replacement exists yet.
Close on the route the decision-maker has actually heard and accepted, paired with the quality it gives up. Where no test has been run, close on the comparison itself—the three routes, the same picture, one afternoon. The take you pitched is still an open question until someone answers it out loud.
Frequently asked questions
What are the three routes for handling a line that cannot be heard clearly?
Repair the original recording, perform the line again, or edit the passage so the damaged line is no longer needed. Each can solve intelligibility but can change something the treatment already sold, so all three should be compared against the same picture.
How can I tell whether the problem is in the recording or in the performance?
Build an inventory first. Identify whether the unintelligible part is one word, a phrase, or the whole line, and ask whether the failure comes from the recording, such as a bad boom position, or from a performer's choice to throw the word away. The second can be expressive and should not automatically be treated as damage.
What should a replacement line be checked against?
Play it against the selected picture and watch timing, physical effort, overlap with the other performer, and the partner's response. If the original line ran under or over another line, the new line has to fit a slot shaped by someone else's timing. A booth read can be perfectly intelligible and still change the apparent intention.
Why can't a treatment promise a repaired or re-recorded line will be indistinguishable from the original?
That is a high bar and cannot be stated before the work is heard. A repaired recording and a re-recorded one can both be good and still be two different takes. The honest treatment names the process and verifies it against the picture rather than certifying an outcome.
Who should approve a changed creative reading?
The person with authority over the take's intent, the same authority that signed off on the performance originally. Repair stays inside sound; a new line introduces a new performance that creative authority should hear; an edit changes picture and should go through whoever approved the cut.