Most automatic video tools fill time with whatever B-roll is nearest: a stock clip of someone typing, a drone shot of a city that isn’t the one being talked about. In a documentary that breaks the spell. When the narrator names a place, a person or a moment, the viewer should be looking at it.
Every shot is a window of words
The narration is recorded as one continuous track that knows when every word is spoken. The film is then cut into shots of a few seconds each, on word boundaries, and each shot’s picture is chosen for the words heard during it. Because the pictures are placed on the same word times as the voice, they can’t drift out of sync.
The order it looks in
- Real video of the thing being said: archive film, newsreels, mission footage.
- Then whichever shows the words best, as equals: a real photograph, motion graphics (maps, charts, timelines, diagrams) or AI animation for moments no camera caught.
- Stock comes last, and only when nothing else fits.
Every pick is checked
Candidates are looked at by a vision model against the exact words of the shot: the right subject, the right era, no watermarks, no near-black frames. AI clips get their own check on frames from across the clip, so bodies stay solid and the action matches the narration. No picture is used twice in a video.
And you have the last word
In the shot editor you can swap any pick for a runner-up, upload your own picture or clip, search again in your own words, or show the line another way. The timing stays locked to the words, and rendering again is free.
It works the same from a topic or from your own script, and it’s what makes Scenyn a faceless YouTube video maker that shows the thing being said.