AI Motion Prompt Guide: 40 Camera and Subject Moves

Short answer: a good motion prompt describes one movement, says what the camera should do, names the details that must stay consistent, and stops. “The subject blinks once; the camera stays fixed; preserve the face, hairline and background” is more useful than a paragraph of cinematic adjectives. Start with a small movement, review the paused frames, and change one instruction at a time.

This guide collects the camera and subject moves we use most often. It is not a list of magic phrases. The same sentence can produce different results on different models, and the source photo matters more than the wording. Use the recipes as a starting point, then adapt them to the image in front of you.

The one-move rule

Every additional movement gives the model another chance to invent something. A blink, a head turn and a camera push are three jobs. If the clip fails, you will not know which one caused the failure.

Start with one movement. If it works, keep the photo and add a second movement in a new run. If it fails, change the photo before you change the prompt. A blurry face or a cropped hand will not be repaired by better adjectives.

Camera moves

Camera moves are usually safer than subject moves because they do not require the model to rebuild a face or a body. They still change the frame, so straight lines and architecture can bend.

Move	Prompt idea	Best for
Slow push-in	“Very slow camera push-in; subject stays still”	Portraits, products, quiet scenes
Lateral drift	“Camera drifts slowly to the right with gentle parallax”	Landscapes, rooms, layered scenes
Slow zoom-out	“Camera slowly zooms out to reveal the surrounding scene”	Products, interiors, character introductions
Fixed camera	“Camera stays completely still”	Blinks, hair, fabric, subtle expression
Low orbit	“Camera orbits slowly around the product at a low angle”	Objects on a clean surface
Tilt up	“Camera tilts slowly upward from the hands to the face”	Portraits with a clear vertical composition
Rack focus	“Focus shifts from the foreground object to the subject”	Products and layered still lifes

Do not combine a push-in with a head turn in the first run. If the face changes, you will not know whether the camera or the head caused it.

Subject moves

Subject moves need visible evidence. A blink needs eyes. A head turn needs a visible jaw and ear. A pet’s ear twitch needs an ear that is not cropped. A fabric move needs folds and light.

Move	Prompt idea	Best for
Blink	“One natural blink; head and camera stay still”	Portraits, pets
Small head turn	“Head turns a few degrees toward the camera”	Portraits
Hair breeze	“A few loose strands move in a gentle breeze”	Portraits, fashion
Fabric shift	“The edge of the clothing moves slowly in the wind”	Fashion, product
Water ripple	“Water ripples gently; shoreline stays fixed”	Landscapes
Cloud drift	“Clouds move slowly across the sky”	Landscapes
Ear twitch	“The pet’s ears twitch slightly; body stays still”	Pet portraits
Slight smile	“A very small, natural smile”	Portraits

If the result looks like a different person, reduce the movement. A smaller motion is easier to keep consistent than a large expression.

Preservation instructions

Tell the model what must not change. Short, specific preservation instructions work better than a long list of generic negatives.

• “Preserve the subject’s face, skin tone and facial proportions.”
• “Keep the same number of people as the uploaded photo.”
• “Keep the original background and camera angle.”
• “Preserve the fabric pattern and color.”
• “Keep the pet’s fur markings and eye color consistent.”
• “Do not add text, logos or watermarks.”

The list should match the failure you have already seen. If you have not seen an extra person, you do not need to list “no extra people” in every prompt. A short list is easier to test.

Negative instructions that help

Negatives are useful when they describe a real failure mode:

• no duplicated fingers or limbs;
• no extra people;
• no lettering or logos;
• no sudden scene cuts;
• no background replacement;
• no anatomy changes;
• no glowing edges or plastic skin.

Do not paste twenty negatives into every prompt. They dilute the instruction and make it harder to know what helped. Add one negative after you see the corresponding failure.

A prompt template

Use this structure for a first test:

1. Source: “Use the uploaded photo as the reference.”
2. Movement: “The subject blinks once and smiles very slightly.”
3. Camera: “The camera stays completely still.”
4. Preserve: “Keep the face, hairline, clothing and background unchanged.”
5. Avoid: “No new people, text or objects.”

This is intentionally boring. Boring prompts are easier to review. Once the basic motion works, add one creative instruction and compare the two clips.

Review the frames

Watch the clip three times. First for the overall motion, second for the face and hands, third for the background. Then pause on the most active frame and inspect the same areas again.

Reject the clip if the eyes change shape, the fingers merge, the hairline moves, the background bends or a new object appears. A clip that only works at thumbnail size is not ready to publish.

FAQ

How long should a motion prompt be?

One or two sentences is usually enough. Describe one movement, the camera behavior and the details to preserve. Longer prompts with many movements are harder to diagnose.

Should I mention the camera and the subject together?

Separate them. “The subject blinks; the camera stays fixed” is easier to test than a sentence that mixes a head turn, a push-in and moving hair.

What negative instructions actually help?

Use a short list for the failures you have seen: no new people, no text, no duplicated limbs, no background changes. Do not paste a generic list of twenty negatives.

Why does the same prompt produce different results?

Image-to-video models are stochastic and depend on the source image, queue, seed and model version. Change one variable at a time and keep a note of what worked.

Try a recipe on your own photo

Open the free animation demo, paste one recipe, and upload a photo you are comfortable sharing. If you want a specific aspect ratio or a different model, compare the signed-in image-to-video tool and the credit packs before you commit.