How to add animated verse callouts to a long lecture video
Upload the lecture and ask Vizard Agent to subtitle it and mark the passages the speaker quotes. Vizard Agent transcribes the whole talk, searches the transcript for references, looks up the real text of each one, and animates a card at the exact moment the quote begins.
What do you type into Vizard Agent?
Ask for both jobs in one instruction. Vizard Agent has to coordinate them, because a card and a subtitle in the same part of the frame at the same moment is the failure case, and asking for both together is what lets it plan the gaps. Name your preferred translation too, since the cards quote whatever version you name.
Prompt
What does Vizard Agent actually do?
It works from the transcript, not from what the transcription heard. Every reference is searched for, the real wording is looked up, and each card lands on the word where the quote begins. Then it checks the finished render instead of shipping it, and on this lecture that check caught a timing error. In the session behind this article it:
- Probed a forty-four-minute lecture, grabbed sample frames, transcribed the whole talk and searched the transcript for every reference the speaker made.
- Looked up the exact text of each passage and read the word-level timings so each card would land on the word the quote begins on.
- Designed an animated card, composited it onto a real frame and diagnosed why the first render came out blank.
- Generated the subtitle file with gaps at the verses, verified no subtitle sat under a card, and compared three subtitle styles on frames: plain, higher-contrast white and a dark backing box.
- Rendered in parallel segments, found the cards had landed at the wrong timestamps, corrected the timing and re-rendered every segment.
- Verified all eleven cards and the subtitles resuming after each one, in a delivered 1920x1080, 30fps file of 2631.73 seconds.
When does this not work well?
Quoting is looser than it sounds. Vizard Agent finds explicit references reliably, but paraphrases, half-quotes and allusions are harder to catch. Translations differ, so name the version or a card may not match the speaker's wording. Book names and numbers are exactly where transcription errors cluster, and forty-four minutes at full quality is a long render, twice over if it needs correcting.
Common questions
Is this only for scripture? No. Any speaker quoting a fixed source, such as case law or poetry, works the same way. Vizard Agent looks up the real wording.
Will the subtitles overlap the cards? No. Vizard Agent builds gaps into the subtitle file and checks frames to confirm.
Can it use my preferred translation? Yes. Name it in your instruction and Vizard Agent uses that wording.