AI Video & Image Prompt Guide

1. Universal prompt formula

aiMakeVideo follows natural language. Stack these building blocks — you do not need every line every time, but clearer briefs usually generate more controllable clips and stills.

2. Pick the right mode

Start from the asset you already have. Switching modes later is fine — Recreate and Create keep briefs in the same studio.

Every example prompt on this page is shown exactly as it was run. Copy it into the studio, or tap Recreate to load it with the matching settings.

Text to Video

Best when the scene only exists as words. Write subject, motion, camera, light, and mood in one brief. Add timed beats for longer clips.

How to write a text-to-video prompt

• Lead with one subject and one action. Name who or what is on screen and the single thing they do. A clip lasts only a few seconds, so a crowded brief gets squeezed and nothing reads clearly.
• Describe exactly one camera move. Pick one move — slow push-in, lateral tracking, orbit or a locked-off tripod — and say how fast it goes. Conflicting moves are the most common cause of jittery footage.
• Add light, setting and what to leave out. Time of day, light direction and texture give the model a look to aim for. Finish with exclusions such as “no text, no subtitles” when a clean frame matters.

Weak prompt

a woman walking on a cliff, beautiful, cinematic

Improved prompt

A woman in a flowing champagne silk dress walks along a cliffside path at golden hour, fabric catching the wind, warm rim light on her hair, slow lateral tracking shot, 85mm look, shallow depth of field, photoreal, no text

Why it works: The improved version pins down the wardrobe, the time of day, one camera move and a lens feel. “Beautiful” and “cinematic” on their own give the model nothing concrete to render.

Try the Text to Video generator

Real examples you can recreate

16:9

Cliffside silk walk

A fashion clip made from text alone: golden-hour light, wind in the fabric and a single tracking move.

Prompt

A graceful woman in a flowing champagne silk dress walking along a cliffside path at golden hour, fabric catching wind, warm rim light on skin and hair, slow lateral tracking shot, 85mm look, shallow depth of field, refined beauty commercial, photoreal, subtle film grain

Sets the wardrobe, the light and one lateral tracking shot, then adds lens and film-grain notes for a commercial finish.

Recreate

1:1

Perfume bottle orbit

A short product loop: the camera circles a bottle while prism light plays across the glass.

Prompt

Luxury perfume bottle on black glass, slow orbit, prism light splitting into soft spectral highlights, ultra-clean reflections, macro commercial beauty, razor-sharp label readable, elegant negative space

A compact product brief — surface, camera orbit, light effect — plus an explicit request to keep the label sharp and readable.

Recreate

Image to Video

Best when you already love a still. Upload or generate the frame first, then describe how it should move while locking face, outfit, and background to the reference.

How to write an image-to-video prompt

• Lock what must not change. Start by saying that the face, outfit, product or background stay as in the reference. The image already defines the look; your words should protect it, not describe it again.
• Spend your words on motion. Describe what moves and how: breathing, hair in the wind, a slow head turn, liquid pouring. Small, physical motion keeps the likeness far better than a big change of pose.
• Script longer clips in beats. For anything longer than a few seconds, split the action into timed beats such as 0–2s and 2–5s. Give each beat one action so the model does not rush everything into the first second.

Weak prompt

make the bottle look cool

Improved prompt

Keep the exact bottle shape, label and colors from the reference. 0–2s: slow push-in toward the bottle as light glides across the glass. 2–5s: gentle low orbit to the right while the petals in the foreground drift softly. Shallow depth of field, luxury commercial, no new text or logos

Why it works: The first prompt leaves the model guessing, so it may redesign the product or invent random motion. The improved one protects the label, gives each part of the clip a single job and rules out new text that could distort the branding.

Animate a photo with Image to Video

Real examples you can recreate

16:9

Reading by the window

A lifestyle still brought to life with subtle, natural motion instead of big action.

Prompt

Soft breeze through the open window. Sheer white curtain and loose hair strands drift gently. She slowly lifts her gaze to look out at the sunlit garden, holds it a moment, then lowers her eyes back to the book. Garden flowers sway softly. Elegant, quiet, cinematic, natural motion only.

Lists only the motion — breeze, curtain, a slow glance up and back down — and ends with “natural motion only” to keep the scene calm.

Recreate

1:1

Oil portrait micro-motion

A classical painting animated so gently you almost miss it: a blink, a breath, the hint of a smile.

Prompt

Nearly still Old Master portrait. Imperceptibly, the corners of her closed lips soften into a faint Mona Lisa smile. One slow, heavy-lidded blink, so slight it is easy to miss. Tiny breath lifts the white chemise a hair. Warm Rembrandt light, painterly texture locked, camera locked. Mysterious, elegant, micro-motion only — if you look away you might think nothing moved.

Locks the painterly texture and the camera, then describes tiny, precise movements so the portrait never loses its likeness.

Recreate

Text to Image

Best for new stills from scratch — product heroes, beauty portraits, lifestyle frames. Specify subject, lens, lighting, and style.

How to write a text-to-image prompt

• Name the kind of image first. Open with the category — product shot, editorial portrait, still life, poster — so the model settles composition and finish before the details arrive.
• Describe light and lens like a photographer. “Soft window light from the left, 50mm, shallow depth of field” does more than any pile of adjectives. Light is what makes a still look premium or flat.
• Choose the size for where it will be used. Set the aspect ratio in the studio before you generate — square for feeds, portrait for stories, wide for banners. Cropping afterwards throws away detail and often cuts through the subject.

Weak prompt

coffee cup photo, high quality, 8k, masterpiece

Improved prompt

Product photo of a matte white ceramic coffee cup on a light oak table, morning window light from the left, soft shadow, thin steam rising, 50mm lens, shallow depth of field, clean minimal background, lifestyle advertising style

Why it works: Tags like “8k” and “masterpiece” add no information. The improved prompt specifies material, surface, light direction and lens — the things that actually decide how the photo looks.

Create stills with the Text to Image generatorCompare image models in the AI Image Maker

Real examples you can recreate

1:1

Hands holding a tea bowl

An intimate still life in which warm side light and natural textures carry the mood.

Prompt

Intimate still life of elegant hands holding a glowing tea bowl, warm side light, linen and wood textures, wabi-sabi beauty, photoreal detail

Names the subject, one light direction and two textures, then a style reference (wabi-sabi) to set the tone.

Recreate

1:1

Pearl earring macro

A jewelry close-up that shows how macro framing and creamy bokeh turn a small product into the hero.

Prompt

Extreme macro of a pearl earring against soft neck skin, creamy bokeh, jewelry beauty advertising, exquisite highlight control

Specifies an extreme macro framing and an advertising context, and asks for careful highlight control on the pearl.

Recreate

Image to Image

Best for restyles and refinements. Say what to keep (identity, logo, product shape) and what to change (wardrobe, lighting, backdrop).

How to write an image-to-image prompt

• Say what to keep, then what to change. Open with what must survive — likeness, pose, product shape, logo — and only then list the changes. Models follow this order more reliably than a description of the new image alone.
• Change one or two things per pass. Relight, restyle the wardrobe or swap the backdrop — not all three at once. Chain several passes when you need a bigger transformation.
• Start from a clean reference. A sharp, evenly lit source with the subject clearly visible gives the model more to preserve. Blurry or heavily filtered photos drift further from the original.

Weak prompt

make it look professional

Improved prompt

Keep the person's face, hairstyle and pose exactly as in the reference. Replace the background with a soft grey studio backdrop, relight with a large softbox from the front left, natural skin texture, business headshot style, no plastic retouching

Why it works: “Professional” can mean anything. The improved prompt protects identity first, then names exactly two changes — backdrop and light — so the result still looks like the same person.

Restyle a photo with Image to Image

Real examples you can recreate

Daylight to candlelight

A relight: the same composition and subject, moved from daytime into a warm candlelit evening.

Prompt

Keep composition and subject. Relight to intimate candlelit beauty: warm key, soft shadows, gentle bloom, romantic evening atmosphere, photoreal continuity

Opens with “keep composition and subject”, then changes only the lighting and mood — a textbook single-change edit.

Recreate

1:1

Product on a luxury set

A product photo placed on a newly styled set while its shape and branding stay intact.

Prompt

Keep product shape and branding. Place onto a soft luxury beauty set with silk, petals, and diffused daylight, premium cosmetic advertising polish

Protects the product shape and branding first, then describes the new set, props and light for an ad-ready image.

Recreate

3. Choosing a model

aiMakeVideo is multi-model. Pick for the job — speed, motion quality, or still detail — then iterate with the same prompt.

4. Common prompt mistakes

Most disappointing results come down to a handful of habits. Check your prompt against this list before you spend credits on another run.

1. Stacking adjectives instead of details. “Stunning, epic, ultra-realistic, masterpiece” tells the model very little. Swap each adjective for something it can actually render: a material, a light source, a lens, a place.
2. Asking for too much in one clip. Three characters, a costume change and a plot twist will not fit into a few seconds of video. Keep one subject and one main action per clip, and join clips afterwards.
3. Giving conflicting camera directions. “Zoom in while orbiting, handheld, drone shot” forces the model to choose, and it often chooses badly. One clearly described move gives smoother footage.
4. Re-describing the reference image. In image-to-video and image-to-image, the picture already carries the look. Repeating it in words can make the model reinterpret it. Say what to keep, then spend your words on what changes.
5. Relying on text inside the frame. Many models render words, logos and captions unreliably. Add text in an editor afterwards, and write “no text, no subtitles” when you need a clean frame.
6. Rewriting everything after one bad result. Change one thing at a time — the camera move, the light or the model — so you can see what made the difference. Test at a short duration and the lowest resolution first, then scale up once the composition works.

5. Credits, ratio & duration

Every job estimates credits before you run it. Cost rises with longer duration, higher resolution, premium models, and extra reference images.