A faceless explainer can make information approachable, but a confident voice and polished image are not evidence. Keep a source note alongside the script and reserve generated imagery for illustration unless it accurately represents an observed event.
Split each beat by evidence type
For a three-minute history of a neighborhood bridge, list each date and structure detail from the records you reviewed. Mark a fictional construction scene as illustrative if you animate it. In the AI video generator, prepare atmosphere shots from original descriptions or appropriately licensed reference material. Use a real archive image where the exact design or historical condition matters.
Keep references out of the voiceover unless they help the narration, and give viewers a usable list in the description. Ensure doc titles identify the specific source rather than a generic homepage.
Check what the visuals imply
Watch the edit with the script turned off. Does an invented crane, roadway or skyline look like evidence of what was there? Replace it with a map graphic, a narrower crop or an explicit label. For a short version, preserve the label near the relevant shot even if the longer introduction was removed.
Then compare the published wording to the original notes. If a claim needs a date, avoid a vague caption such as recently. Viewers should be able to understand where facts end and the demonstration begins.
If a voice tool reads a place name incorrectly, verify the pronunciation from the relevant community or speaker. Keep the official spelling in the captions and source list, which people may copy for their own research.


