How to catch a caption card that claims something the clip never said

Last updated 2026-09-15 · 8 min read

What is the short version?

A pull-quote card is a summary, and summarising a single line out of a long conversation means guessing at the context that line came from. Guess wrong and your video states, in large type and with complete confidence, something the speaker never actually said.

1. Go to Vizard Agent with the cut.
2. Say which card is wrong and what it wrongly claims.
3. Ask it to re-read the context around that quote.

What do you need before you start?

The clip and the truth. Vizard Agent can re-read the transcript, but where the correct version depends on knowledge outside the recording — who someone is, what their title is, whose story it was — you have to supply it.

• The cut. With the cards in it.
• What the card wrongly claims. In your words.
• The actual situation. If it is not in the clip.
• Roles and titles. For anyone named.
• The full recording. Context often sits outside the cut.

What do you type into Vizard Agent?

Say what is untrue, not what reads badly. "That card is wrong" invites a rewrite for style; naming the claim tells it a fact is at stake, which sends it back to the transcript rather than to the thesaurus.

Prompt

Variants worth knowing:

• "It says X — it was not X." Names the claim.
• "Re-read the context around that quote." The method.
• "Only what the clip supports." The standard.

What does Vizard Agent actually do?

Here is the sequence Vizard Agent worked through on a real podcast teaser built around typographic cards. The cards were designed, reviewed and approved for how they looked several times over before anybody stopped to check what they were actually asserting.

1. Reads the transcript to find the lines worth pulling.
2. Builds the cards from those lines.
3. Reviews them for design and timing.
4. Is told a card states something untrue.
5. Re-reads the story context around that quote.
6. Rewrites the card to what the clip supports.
7. Corrects anybody's title or role on screen.
8. Rebuilds the timeline with the corrected cards.
9. Re-checks the other cards against their own context.

Step five is the whole method. A quote is a sentence; the claim a card makes is about the situation the sentence came from, and only the surrounding lines establish whether that situation is what you assumed.

Step nine is the part that saves the next round. One card written from too little context usually means the others were written the same way, so they get checked rather than left until somebody notices a second one.

Step seven is a smaller version of the same error. A person labelled by what they appear to be rather than by what they are is a factual claim on screen, and it is exactly as wrong as a misattributed story.

What does the result look like?

Cards that say what the clip actually says. They are often a little less punchy than the wrong version was, because the wrong version carried the confidence of an assumption inside it, and that is the trade you are knowingly making here.

Vizard Agent also tells you which lines each card was drawn from, so you can check its reading of the conversation against your own.

When does this not work well?

Some facts are simply not in the recording. If the clip never establishes who somebody is or whose story is being told, re-reading the transcript will not produce the answer and the card has to be written from what you know.

• Facts outside the recording. Only you have them.
• Deliberately ambiguous speech. The speaker did not say.
• Heavily cut clips. The context was removed.
• Titles and roles. Rarely stated aloud.
• Claims about third parties. Check before asserting them.

How do you fix a result that came back wrong?

State the correct fact rather than suggesting better wording. Because Vizard Agent writes each card out of a fact rather than out of a phrase, handing it the right fact repairs that card and every other graphic that was built on the same wrong assumption.

• "It was his friend's daughter." The card and the rest follow.
• "His title is different." Corrected wherever it appears.
• "The card is right but flat." Now it is a wording note.
• "Check the others too." All cards against their context.

How does Vizard Agent compare to doing it yourself?

By hand a card is written from the line that was pulled, because the line is what is in front of you in the edit. The context is two minutes earlier in a recording nobody reopens, and the assumption goes out in large type.

Common questions

Why do cards get facts wrong? Because a single line rarely carries the situation it came from.

Can it check every card? Yes. Ask Vizard Agent to verify each one against its context.

What if the fact is not in the clip? Then Vizard Agent will say so and ask you for it.

Will the card be less punchy? Sometimes. Accuracy costs a little confidence.

Does it check titles too? Yes, though Vizard Agent needs the correct title from you if nobody says it aloud.

Can it quote exactly instead? Yes, and Vizard Agent will note that a verbatim quote makes no claim of its own.

What about the full recording? Give it to Vizard Agent; the context is often outside the cut.

Does this apply to thumbnails? Yes. Vizard Agent treats a thumbnail as making a claim the same way a card does.

Will it flag risky cards itself? Yes. Vizard Agent tells you which ones rest on inference rather than on the words.

Can it show me the source lines? Yes. Vizard Agent lists them so you can compare its reading against yours.

What if two readings are possible? Vizard Agent will say so rather than choosing one silently.

Does the fix reach the captions? Captions carry what was said, so Vizard Agent usually finds them already correct.

Can it rewrite without losing the hook? Often. Ask Vizard Agent for options that stay inside the facts.

Why was it approved the first time? Because it was reviewed for how it looked, not for what it said.