Direct Answer
Edit an interview clip without changing meaning by locking the source before cutting. Save the full answer, verify the transcript, write the claim and caveat, then make an edit that preserves subject, evidence, certainty, chronology, and intent. Review the final clip against the source and get speaker approval for sensitive or consequential material.
Preserve meaning rather than every pause: a reasonable viewer should understand the same proposition the speaker intended in the full conversation.
Step One: Create a Source Lock
Before opening the timeline, save the full recording, the question that prompted the answer, the checked transcript, start and end timestamps for the full answer, the one-sentence claim, supporting source links, the material caveat, the speaker and role, and permission status.
Do not let later exports replace the source. Give the recording a stable ID.
The Society of Professional Journalists' Code of Ethics is written for journalism, not brand content, but its principles of accuracy, context, source clarity, and correction provide a strong editorial baseline.
Step Two: Mark the Semantic Spine
Highlight the minimum parts the answer needs: subject, claim, basis, scope, caveat, and consequence.
Subject means who or what the speaker means. Claim means what the speaker says is true or advisable. Basis means the example, evidence, or experience. Scope means the time, group, condition, or role. Caveat means what the claim does not establish. Consequence means why the audience should care.
These elements do not need to appear in that order, but the final cut must not lose a part that changes interpretation.
Step Three: Choose In and Out Points by Meaning
Begin before the subject becomes clear. End after the consequence or caveat resolves.
A weak start is, "which is why we removed it." A stronger start is, "After six onboarding interviews, we removed the required kickoff call for experienced teams."
A weak end is, "Conversion improved." A stronger end is, "Trial-to-paid conversion improved in that cohort, but we have not tested the change with first-time teams."
A few extra seconds can prevent a much larger misrepresentation.
Step Four: Clean Without Reconstructing
Usually safe edits include removing long pauses, removing duplicate false starts, cutting unrelated tangents, tightening repeated setup, removing filler that does not affect tone or certainty, and reordering visual B-roll while the continuous audio remains intact.
Edits that need explicit review include joining words from separate sentences, reordering sentences, removing hedges or qualifications, converting a question into a declarative setup, combining answers from different questions, using reaction shots from another moment, and changing the apparent sequence of events.
If a sentence must be reconstructed, document the change and let the speaker approve it. Never splice language to create a statement the person did not make.
Step Five: Verify Captions Against Audio
Captions are part of the claim. Automatic captions can change names, figures, negations, and technical terms.
Check proper nouns, numbers and units, "can" versus "cannot," dates, acronyms, speaker labels, punctuation that changes meaning, and words hidden by music or compression.
The W3C explains why captions and transcripts matter for accessibility. Accurate captions also protect meaning for viewers watching without sound.
Do not replace the spoken caveat with a shorter caption that removes it.
Step Six: Run the Title and Thumbnail Test
The packaging must stay within the clip.
If the clip says, "For our first ten design partners, a manual review caught errors automation missed," then "Manual Review Always Beats AI" is misleading.
"Why We Kept Manual Review for Our First Design Partners" is a more faithful title.
A title can frame the question. It cannot enlarge the evidence.
Step Seven: Compare Source and Cut
Play the full answer, then the final clip.
Ask whether the claim is equally certain in both, the subject is the same, the time period is preserved, a hypothetical is now presented as fact, an estimate is now exact, a joke has become advice, disagreement has disappeared, the speaker would recognize the meaning, and a reasonable critic would call the cut fair.
Record reviewer and approval date.
A Redline Example
Full answer: "We think the new queue reduced response time because median first response fell from nine hours to six during the two-week pilot. The pilot also had lower volume, so we cannot attribute the change to the queue yet."
Misleading edit: "The new queue reduced response time from nine hours to six."
Faithful edit: "During the two-week pilot, median first response fell from nine hours to six. Volume was also lower, so we cannot attribute the change to the queue yet."
The faithful version preserves the observation and removes an unsupported causal conclusion.
Preserve an Edit Decision Log
For consequential clips, save removed sections, reordered material, caption corrections, title rationale, disclosure, reviewer, speaker approval, published URL, and correction or retirement history.
Technical provenance standards such as C2PA can record aspects of an asset's history. An internal edit log is not the same as a Content Credential, but it supports accountability.
Honest Limitations
Faithful editing cannot make an unsupported answer true. It cannot solve missing consent, copyrighted footage, unsafe advice, or a source that should not have been recorded.
Some answers need too much context for short video; publish a longer excerpt or written explanation.
Speaker approval can also miss harm to other people named in the story. Add legal, privacy, or subject review where risk requires it.
How REC Fits
REC Content Studio researches supplied context, prepares a guided solo interview, records the person's answer, creates a transcript, and identifies candidate highlights. The editor still controls the cut and final representation.
Use the full REC recording and question as the source lock. Treat the transcript as a navigation aid, not infallible text. Cut time, not meaning; then keep the decision trail beside the asset.