NARRATION AND CARDS
Hypit narration cards: change the picture when the words arrive
Watch Daojie’s Chinese adaptation comparison and map script lines to example cards, list highlights and captions.
Edited by HypitAI · Reviewed
Watch lists, examples and captions change together
Daojie’s roughly 50-second screen recording compares reference and Chinese versions. The accompanying post describes a local production workflow.
View Daojie and the original X post →Choose three examples of your own
Use this format for a tutorial list, product steps or an explainer. Highlight the current item while a card supplies the corresponding picture. The desk-organisation example below is our filled production brief.
Map every line to a card
Put narration, assets and transition conditions in one table. Refer to spoken words rather than guessing timestamps before the voice is ready.
| Narration | Trigger phrase | Asset | Display and end condition |
|---|---|---|---|
| Divide the desk into work and storage zones. | work | cards/01-zones.jpg | Highlight item one; hold until item two |
| Keep daily items within reach. | daily items | cards/02-daily.jpg | Highlight item two and show frequently used objects |
| Put occasional items in one box. | one box | cards/03-box.jpg | Highlight item three and show the box |
| Check that there is still clear space. | Check | cards/04-finish.jpg | Show the finished desk through the ending |
Prepare the files
Supply script.txt, approved voice.wav, a cards/ image directory and cards.csv. Include item IDs, lines, trigger phrases, assets and end conditions. An optional small presenter also needs a character image and a position that does not cover captions or cards.
An editable narration-card brief
This is our example brief, not a prompt published by the creator. Replace the files and subject with your own.
/hypit Make a vertical desk-organisation explainer using script.txt, voice.wav and cards.csv.
Keep the full voice track. Show a list and highlight its current item.
Switch cards/ images at the spoken trigger phrases in cards.csv rather than estimated timestamps.
Align the actual voice first, then preview the line-to-card mapping.
Hold each card until the next item; hold the final card through the end of the audio. Do not leave an empty frame.
Use readable sentence captions without covering the presenter or example.
Use the supplied assets for the first version and report missing files.
Deliver the editable project, MP4 and card mapping.Approve speech before timing the cards
The official timing workflow aligns the script with recorded speech and maps words to visual timing. Check missing words, pronunciation and pauses first. A new voice or script requires alignment and caption/card review for the affected sections.
Review the complete MP4
Check the output while listening, not only the static preview.
- Does the list highlight move with the new example card?
- Is each card readable, and does a finished video card leave an empty box?
- Do captions follow the voice without overlapping at sentence boundaries?
- Can the presenter, cards and captions all be read on a phone?
Treat old workarounds as historical
Daojie’s post records frame-rate and AAC sample-count issues. Official 0.2.6 documents caption-boundary and final audio-mux fixes. Use the current troubleshooting guide rather than copying old FFmpeg downgrade or codec-replacement scripts.