Section 5: Synthetic Voice and Audio Direction · Lesson 4 of 5 (24 of 45) · 60 minutes · Evidence: Clean pickup package
Think about a film or video where one tiny audio flaw pulled you out of the story: a name said wrongly, a sudden hiss, a line louder than the one before it. Audiences forgive a rough picture far longer than they forgive rough sound. Today you learn the repair kit: fixing a single line without redoing the whole take, and making all your lines sit evenly together.
Production journal: write your answer below.
What must a filmmaker understand before they can correct names and difficult words record pickups trim silence and balance level?
Most voice problems fall into four families, and naming the family tells you the repair. First, pronunciation: a name, a place or a number is said wrongly. Second, timing: a pause is too long, a line is rushed, or the voice enters before the picture is ready. Third, noise: breaths, mouth clicks, hiss, hum, room echo or background sound that is not meant to be there. Fourth, level: the line is too quiet or too loud compared with its neighbors.
A pickup is a short re-recording or regeneration of just the flawed line or phrase. You do not rebuild the whole scene. Keep the same authorized voice and the same direction note, correct the single problem (for example, a respelled pronunciation cue), and generate or record the pickup. Then place it where the flawed line was. Always give the pickup its own file name and keep the old one, so a teacher can see what changed.
To trim is to cut silence, a stray breath or a false start from the beginning or end of a clip, leaving a natural pause rather than none: audiences need a beat of air. Noise is unwanted sound. Removing a loud breath or click is a cut; steady hiss may need a noise-reduction function in your editor, applied gently, because heavy reduction makes a voice sound watery. Level is how strong a signal is, and loudness is how strong it sounds to the ear. Your goal is consistent loudness: lines in the same scene should sit at a similar volume, so the audience never reaches for the remote. Many editors, including CapCut, offer a normalize or loudness-matching function; use it to even out lines, then listen on speakers and headphones and adjust by ear. Do not push any line so loud that it distorts.
A pickup must still sound like the same character. If your repair switches to a different voice, or the pickup is so different in tone that it sounds like another person, you have broken continuity and, if the new voice is not on your casting record, you may have broken your authorization too. Match the pace and tone of the neighboring lines.
In Last Light, the voice says the place name as ‘Cair Dor-ann’, but it should be ‘care DOR-an’. The student diagnoses a pronunciation problem and writes a respelled cue. She regenerates only that sentence as a pickup named with the line ID and ‘pickup1’. She trims half a second of lead-in silence, cuts a loud breath before the next line, and normalizes the two lines so they match. She listens through the scene, finds the pickup a touch brighter, and slows its pace slightly. The corrected line goes in; the old file stays in the archive.
Dialogue editors, whose entire job is cleaning and matching speech, work through a steady sequence. Study it as if you were looking over one editor’s shoulder, and pause at each decision to ask who is making the call: the person, or the tool.
The editor listens through the dialogue first without touching anything, and writes a list of problems with a time mark for each. They sort the list: which are pronunciation, which are timing, which are noise, which are level. They fix the problems that need new audio first, by requesting pickups, because trimming a line that will later be replaced wastes effort. Only then do they trim, remove noise and balance level across the scene.
Two habits mark a careful editor. They compare before and after, switching quickly between the original and the fix, to make sure the repair has not damaged the voice. And they check the sound on more than one kind of playback. The tool can do the cutting and the matching in a click, but it cannot hear whether the result sounds natural.
As you study, record each decision point in your production journal.
| Decision point | What the filmmaker does | What the tool automates | What still needs human review |
|---|---|---|---|
| List the problems | Listens without editing and logs each problem with a time mark | Nothing; this needs a trained ear | Whether a problem is a real flaw or a deliberate choice, such as a hesitation |
| Request pickups | Respells the cue, keeps the voice and direction, and requests just the flawed line | Generates or records the replacement | Whether the pickup matches neighboring lines in pace and tone |
| Trim and clean | Cuts stray silence and loud breaths and applies gentle noise reduction | Applies the cut or the reduction function | Whether the pause still feels natural and the voice is not watery |
| Balance loudness | Matches line loudness across the scene and checks on speakers and headphones | Normalizes levels on request | Whether the lines feel even to a listener and nothing distorts |
Studio task: Diagnose pronunciation timing noise and level problems in sample clips.
Your teacher provides sample clips with deliberate flaws, or use a short sequence from your own film’s takes. Work with a partner as dialogue editors.
Coaching tips. If you cannot decide whether something is a flaw, ask whether it serves the character: a hesitation before bad news may be exactly right. When you apply noise reduction, start gently and increase only until the problem fades; if the voice starts to sound underwater, you have gone too far. If two lines sound unmatched in level even after normalizing, check whether they were recorded or generated with different settings.
Clip/line ID: ______ Time mark: ______ Problem family (pronunciation / timing / noise / level): ______ What I heard: ______ Repair (pickup, trim, noise reduction, loudness match): ______ Respelled cue used (if any): ______ Same voice and direction note kept? yes / no Original file: ______ Replacement file: ______ Before/after check on speakers and headphones: ______
Studio task: Create corrected pickups and a clean replacement sequence.
Build a clean pickup package for your own film.
Checklist before you submit:
List the file name of the approved pickup.
Deliverable: Clean pickup package.
Submit the clean pickup package. The file must use the course naming convention and include enough process evidence to show the student's decisions and revisions.
File name (course naming convention): AIF_S05-L04_clean-pickup-package_v01_YOURINITIALS
| Criterion | Advanced | Proficient | Developing | Beginning |
|---|---|---|---|---|
| The work addresses the stated story or production purpose. | Every repair is chosen because it serves a specific scene, and the sequence plays as a seamless piece of the film. | Repairs serve the scenes and the sequence plays smoothly. | Repairs are made but not clearly tied to the scenes. | Repairs do not connect to the film’s purpose. |
| The student applies the required method accurately and produces usable capstone evidence. | Diagnosis, pickups, trims, noise reduction and loudness matching are all applied correctly, in a sensible order, with accurate file naming. | Pickups, trims and loudness matching are applied correctly and labeled. | Some repairs are missing or applied unevenly. | Little or no technical repair evidence. |
| The student identifies at least one quality issue and records a meaningful revision decision. | Before-and-after listening exposes a real remaining flaw and records a specific second revision. | Records at least one quality issue and a meaningful revision decision. | Notes a flaw without a clear decision. | No revision decision recorded. |
| The work complies with consent rights attribution privacy disclosure and school policy. | The approved voice identity is preserved in every pickup, originals are kept, and the authorization record stays accurate. | The approved voice is preserved and originals are kept. | The voice drifts in a pickup, or originals are not all kept. | A pickup uses an unapproved voice or overwrites originals. |
Requirement: Preserve the approved voice identity while correcting performance.
Correcting a performance does not give you permission to change whose voice it is. Keep every pickup in the same authorized voice that appears on your casting record, and use the same permitted settings. If a consenting speaker recorded the original, ask them before changing a line’s meaning or adding words they did not say, and never use cleanup tools to make a person say something they did not agree to say. Keep the original files, so the history of your edits can be shown.