Ask someone to prompt a video edit and you get a paragraph of vibes. Make it look cinematic. Add some motion. Punch it up. The model has no idea what any of that means, so it produces something generic and the person concludes AI video editing is overhyped.
Ask a working editor the same question and you get a name. Match cut. Speed ramp. L-cut. J-cut. Whip pan. Those are not adjectives, they are instructions with a known result attached. The editor is not better at prompting. They have a vocabulary.
That vocabulary is the whole game, and it is learnable in an afternoon.
You cannot prompt what you cannot name
Every effect in a timeline is a decision someone made: this transition, this easing, this grade, this caption treatment. In a manual workflow you learn those decisions by watching other people’s edits and reverse-engineering them. That takes years, and most of it is spent learning the names.
Skip the years. Learn the names first, then let the model handle the execution. A named effect is a search term, a prompt, and a quality bar all at once. It tells the model what to build and it tells you whether what came back was right.
What follows is twenty names organized by the job each one does. Not a list of features. A list of jobs, because that is how you will actually reach for them.
Captions and type treatments
Text is the highest-leverage element in short-form video, and the treatment you choose signals the entire genre before a word is read.
- Knockout captions — large type that punches through the background with a cutout mask, so the subject sits inside the letterforms
- Serif stamp — a heavy serif line that lands hard with no easing, reading as an editorial seal
- Icon strike — a glyph that hits at the exact frame of a word for emphasis, usually with a small scale punch
- Keycaps — keyboard-style labels that press down to register a beat or a step change
- Comment drop — a screenshot of a real comment sliding into frame as social proof
Each one is a different register. Knockout captions read as designed. Serif stamp reads as authoritative. Keycaps read as instructional. Pick by what the video is doing, not by what looks impressive.
Motion and transitions
Motion is punctuation. Most amateur edits fail here because everything moves at the same speed, which is the visual equivalent of writing without a period.
- Picture dive — a push into a still image until it resolves into the next shot
- Real cascade — layered elements falling into place with staggered offsets, not simultaneous
- Playing grid — a contact-sheet of clips that animates in and selects one
- Video card — a framed clip that slides in as a physical object within the scene
- Lens ring — a circular wipe through a lens element, useful for scene resets
The tell of a professional transition is that it motivates itself. It moves because something in the content asked it to, not because the editor needed a cut point.
Frames and layout
These effects borrow a familiar form so the viewer instantly understands the register. You are not inventing a visual language. You are invoking one.
- Magazine cover — masthead, cover lines, issue furniture wrapped around your subject
- Screenplay page — courier type on white, slug lines and dialogue as a narrative device
- Contact sheet hook — a grid of thumbnails that implies volume and choice in the first two seconds
- NLE timeline — an editing interface rendered on screen, which reads as behind the scenes
- Prompt bar — a typing input field that shows the instruction as it is entered
The NLE timeline and the prompt bar are the most meta of the set, and they work precisely because the audience recognizes them as process. Showing the work has become its own aesthetic.
Look and grade
Grade is the fastest way to make generated footage feel deliberate rather than found.
- Film look — halation, grain, slight gate weave, compressed highlights
- Grade collapse — a transition where the color grade shifts dramatically to mark a time or tone change
- Grease pencil sheet — hand-drawn animation overlay that reads as a storyboard coming to life
Utility
Two workhorse effects that do jobs rather than make statements.
- Live markup — drawn annotations appearing over footage in real time, as if explained as you watch
- Jiggle delete — a quick shake that masks a jump cut so an edit disappears
Before prompting, decide what the effect has to accomplish: prove volume, mark a time jump, add authority. Then pick the name that does that job.
“Add knockout captions to the first line” beats “add some text effects.” Vocabulary is the prompt.
The name gets you the right effect. The number gets you the right feel. Both are required.
If it needs three rounds of revision, the original name was probably wrong for the job.
Use prompt-driven effects when the effect is a known pattern and your constraint is time: captions, transitions, grades, frame treatments. Open a real editor when the effect has to match footage you shot, when timing has to sync to a music bed beat by beat, or when the edit is the creative idea itself. Prompting is fast at executing familiar moves and slow at inventing new ones.
Build the vocabulary before you build the reel
You will not learn twenty effects by reading them once. You will learn them by needing one, going to find its name, and using it. Start with the three your content type needs most and add names only when you hit a job you cannot describe.
The goal is not a larger vocabulary for its own sake. It is that the moment you know something looks wrong, you can name the fix and ask for it in one line instead of three rounds of “a bit more cinematic.”