This guide covers the moves that survive a short generation, the five lines every motion prompt needs, and the words that quietly wreck a shot. It is written for Create video, where you pick a frame and press once per take.
The prompt moves the frame; it does not redraw it
By the time you reach Create video, the frame exists. The model has one job: continue that image forward in time. A prompt that describes a character, a costume, or a style is mostly ignored — or worse, it competes with the frame and the image drifts.
Two things the prompt does control: where the camera goes, and how small things inside the frame move. That is the whole list. If the face is wrong, change the reference image. If the pose is wrong, regenerate the frame on Create. No amount of motion wording fixes either of those.
Six camera moves that work
These are the six moves that behave predictably in a 5- or 10-second shot. Pick one per take.
- Slow push in. The default. Adds attention without changing the composition. Best on a face, a held expression, or a reveal.
- Slow pull back. Reveals the space around the subject. Works when the frame already contains a place worth showing.
- Short orbit. A few degrees around the subject reads as depth. Large arcs usually warp the background.
- Steady pan. Left or right across a fixed scene. Use it when the subject is still and you want the eye to travel.
- Handheld drift. A small breathing motion, not shake. Gives urgency without smearing the face.
- Hold still. The camera locked, only ambience and a tiny subject motion. The safest move, and the one people underuse.
ONE CAMERA LINE, WRITTEN PROPERLY
The camera pushes in slowly and steadily, keeping the subject centred.
Write five lines, not a paragraph
The most reliable motion prompts are short and ordered. Five short lines cover camera, subject, pacing, atmosphere, and the failure you want the model to skip.
Camera move
Name one move and one speed: “the camera pushes in slowly and steadily.” The word slowly is doing most of the work.
Subject motion
One small physical action: “her hair lifts in the wind; her shoulders rise with a breath.” If the action needs choreography, the shot is too long.
Pacing
“One continuous take, no cuts.” This stops the model from inventing an edit between two shots that never happened.
Ambience
Something moves besides the subject: “dust drifts through the light.” Without it, a locked-off frame looks frozen.
Avoid
“No morphing, no new characters, no text.” Name the failure you have actually seen.
A COMPLETE MOTION PROMPT
The camera pushes in slowly and steadily. Her hair lifts in the wind while her shoulders rise with a breath. Dust drifts through the light; one continuous take, no cuts, no morphing, no text.
Match the move to the shot length
A 5-second shot costs 45 credits and usually takes about 3–6 minutes; a 10-second shot costs 90 credits and usually takes about 6–12 minutes. That is twice the runtime at twice the price, and it is a separate, slower generation — not two short clips joined.
Duration changes what a move looks like. “Slow push in” at five seconds is a nudge; the same words at ten seconds are a full arrival. If you want more movement, add time rather than adding a second camera instruction.
Speed is the usual failure. A fast zoom or a whip pan is asked to cross the whole frame in a few seconds, and the model overshoots: edges stretch, the background tears, the face goes soft. Slow is not a style choice here; it is what the generation can actually hold.
Words that wreck a take
Most bad takes contain one of these lines. Read the prompt before you blame the frame.
If the motion is wrong, the prompt is the first place to look. If the face is wrong, the reference is the first place to look. Those two problems have different fixes, and swapping them wastes takes.
- Two moves at once: “zoom in while orbiting and panning.” Pick one.
- Cut language: “then cut to”, “montage”, “scene change.” The model tries to honour it and breaks the frame.
- Transformation: “turns into”, “morphs”, “transforms.” Image-to-video continues the frame; it does not swap the subject.
- Vague intensity: “dynamic”, “cinematic”, “epic.” They push toward fast, unstable motion. Say slow, steady, or a number of degrees.
- Style words: “anime style”, “4K”, “masterpiece.” The frame already fixed the look; extra style words fight it.
- New content: “add rain”, “another character walks in”, “text appears.” There is no clean way to introduce it mid-shot.
Debug a take with one change at a time
When a result is close but not right, change one variable and press Generate again. Pressing twice with two changes tells you nothing.
- Too much motion: shorten the camera line, keep the subject action.
- Face drifts: replace the reference with a closer, front-lit frame. “Keep the face” in the prompt does nothing.
- Frame warps: remove the second camera move, or switch to the longer duration.
- Looks frozen: keep the camera still, add ambience, and give the subject one small action.
- Muddy detail: simplify the background and slow everything down.
BEFORE AND AFTER A DRIFT FIX
Before: “fast zoom in on her face.” After: “the camera pushes in slowly and steadily; her expression stays calm; no morphing.”
The two-minute checklist
Before you press, six things should be true:
If a take still surprises you, the beginner guide covers the reference-image rules and the fix for each common failure.
- One frame, one subject, matching aspect ratio.
- Exactly one camera move, with a speed.
- One small subject motion.
- One ambience detail.
- One avoid line.
- A duration chosen on purpose.
A TAKE THAT CHECKS ALL SIX
One frame of a single character. The camera pushes in slowly and steadily. Her shoulders rise with a breath; dust drifts in the light. No morphing, no cuts, no text.
Questions
Common questions
- Does the motion prompt change the character?
- No. The reference image decides the face; the prompt decides movement. If the character looks wrong, fix the reference, not the wording.
- Can I reuse the same motion prompt on a different frame?
- Yes, with a caveat. A push in or a hold works on most frames; an orbit or a pull back depends on what surrounds the subject. Check that the composition still makes sense before you press.
- Why does my camera shake when I did not ask for it?
- Usually “handheld” or “cinematic” is in the prompt, or two moves are fighting each other. Remove both and ask for a slow, steady move.
- How long should an image-to-video prompt be?
- One to three sentences, around thirty to sixty words. Longer prompts with five camera instructions do not add control; they add conflicts the model has to resolve somehow.
- Should I mention anime style in the motion prompt?
- No. The frame already fixed the style. Style words in the motion prompt compete with the reference image and tend to make the motion less stable.
Try the workflow on one frame.
New devices start with free credits; nothing spends without your press.
FicReel is an AI anime image and video studio. Prices and wait times on this page come from the same credit table the product uses; if the product disagrees with this page, that is a bug — write to admin@mail.ficreel.com. More: beginner guide, pricing, all guides, glossary, about FicReel. Canonical source: https://ficreel.com/learn/image-to-video-prompt-guide.