- PrepareRefine the script, gather useful photos or clips, and set the voice and visual style.
- AssembleCreate an editable video, then review the narration, scenes, captions, and pacing as one story.
- FinishCorrect meaning first, refine the timeline, and run a final publishing check before export.
Prepare the script for speech.
Write for the ear, not the page. Short sentences are easier to narrate, caption, and match with visuals. Each sentence should introduce one clear idea before moving to the next.
Read the script aloud before creating anything. This reveals long phrases, awkward transitions, and words that look natural in writing but sound unclear when spoken.
- Open with enough context to make sense on its own.
- Check names, specialist terms, dates, and numbers.
- Keep production notes separate from the spoken narration.
- Remove any sentence that does not move the story forward.

Give every line a visual purpose.
Go through the script one sentence at a time and ask what the viewer should understand at that moment. A useful visual identifies a person, establishes a place, demonstrates a process, shows a change, or supports a factual statement.
Upload the strongest relevant photos or clips first. You do not need a separate file for every sentence, but you need enough variety to avoid repeating one image without a reason.
For a more detailed editorial process, read how to choose source material for a news video.

Choose only the settings that change the story.
Choose a voice that fits the subject and intended pace. A technical explanation may benefit from an even delivery, while a personal story may need something warmer and less formal.
An AI avatar is optional. Use one when a recurring guide helps connect the opening, transitions, or conclusion. The subject matter should still carry the main visual explanation.
Keep the AI instructions specific. Describe the result you need rather than adding broad style words that do not change the story. If the narration still needs work, follow the explainer-script structure.

Review the generated video as a complete story.
Select Create video when the script, media, and main creative choices are ready. Shortsly assembles the narration, scenes, captions, motion, and sound on an editable timeline.
Treat the generated video as a starting point and review it in four passes:
- Watch without pausing. Is the story easy to follow?
- Listen closely. Does the voice sound natural?
- Read the captions. Are the words and emphasis accurate?
- Pause on each scene. Does the visual support the spoken line?
See the same process across three productions: the Taipei 101 explainer, the FIFA news-video workflow, and the script-only story workflow.

Correct meaning before adding polish.
Start with any scene that is factually wrong, misleading, or unrelated to the narration. Then correct the voiceover and captions, followed by timing and visual polish.
- Replace inaccurate or weak visuals.
- Correct caption text, line breaks, and emphasis.
- Adjust clip positions and scene timing.
- Refine motion, transitions, music, and sound effects.
- Preview the complete sequence again.
A simple edit with accurate visuals is stronger than an elaborate edit that makes the story harder to understand.

Run one final publishing check.
Watch the finished video once with sound and once muted. The first pass checks narration and rhythm; the second reveals whether the visual sequence and captions can carry the story on their own.
- Verify names, dates, numbers, and quotations.
- Confirm that every uploaded file can be used for the intended publication.
- Check caption contrast over every background.
- Keep important text and faces clear in the vertical frame.
- Make sure the final scene does not end before the last word finishes.
When the review is complete, export the finished video.
One final pass with sound. One without.
- FactsNames, dates, numbers, quotations, and source context are accurate.
- FrameFaces, captions, and important details remain clear in the vertical crop.
- RightsEvery uploaded file can be used for the intended publication.
- EndingThe final scene holds long enough for the last word to land.


