AI Video & Image Prompt Guide

Write better prompts, pick the right mode and model, then recreate from real showcase examples.

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.

Subject

Who or what is on screen — person, product, animal, or scene hero. Keep identity consistent if you will iterate.

Action / motion

What happens over time: walk, pour, turn, orbit, wind in fabric. For stills, imply pose and micro-gesture instead of timeline.

Camera

Shot size, lens feel, and moves — push-in, orbit, handheld follow, locked tripod. One clear move beats three conflicting ones.

Lighting

Key light direction and quality — soft beauty, hard noon, golden hour, neon rim, studio softbox.

Mood & style

Overall look: photoreal commercial, editorial beauty, cinematic thriller, quiet lifestyle. Say what to avoid (no text, no extra people) when it matters.

Template

[Subject] + [action] + [camera] + [lighting] + [environment] + [mood]. Optional: timing beats (0–2s / 2–5s) for video.

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.

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.

Open Text to Video

16:9

Volcano Forge Escape

Photoreal live-action fantasy escape through an active volcanic forge cavern, crisp high-resolution cinema detail, ember heat haze kept optically sharp, clean HDR contrast, one wide lens and one uninterrupted controlled handheld pursuit shot. The smith-mage is a muscular North African woman, late 20s, shaved sides with dense curls, glowing ember freckles, soot-streaked bronze armor over ash linen, one living molten hammer in her right hand. Keep one identical smith-mage throughout. Exactly 5 distinct lava constructs emerge; construct 5 is a towering slag titan with a chain-anvil. Never reuse a construct. Every hammer blow hits exactly one target. Show 5 separate contacts, cooled-glass shatters, freezes into dead stone. No multi-target smash. Each next construct blocks, hurls slag, or collapses a bridge. Rotate overhead forge strike, lateral quench-smash, vault over magma rivulet, ground-spike kick, and final titan knee break. Every relocation: micro-anticipation; orange-white forge flash hides her then reveals her already striking the next construct; camera keeps physical follow. No transit smear, no whip pan, no motion blur. Keep skin, armor, lava crust, particles sharp. No dialogue or subtitles. Furnace roar, 5 quench-impact accents, rock collapses, flash cracks, industrial percussion sting.

16:9

Maglev Train Duel

Photoreal live-action modern wuxia duel atop a speeding maglev at golden hour, crisp high-resolution cinema detail, wind-whipped coat fabric, clean HDR contrast, one 40mm-equivalent lens and one uninterrupted controlled handheld tracking shot along the train roof. The duelist is a lean East Asian man, early 30s, sharp jaw, windblown black hair tied low, charcoal tailored coat over combat layers, one carbon-fiber jian in his right hand. Keep one identical duelist throughout. Exactly 6 distinct masked rivals climb onto the roof; rival 6 is a twin-blade captain. Never reuse a rival. Every cut lands on exactly one opponent. Show 6 separate contacts and individual falls off-frame. No multi-target slash. Each next rival lunges, binds, or forces the roof edge. Rotate descending cut, rising cut, bind-and-displace, reverse-grip draw, spinning deflection, and final disarm. Every relocation: micro-anticipation; pale steel flash hides him then reveals him already committed beside the next rival; camera keeps rolling with train landscape visible. No transit smear, no whip pan, no motion blur. Camera stays close with sidesteps and arcs toward each flash destination without snapping. No dialogue or subtitles. Wind roar, rail hum, 6 blade accents, dry flash cracks, accelerating taiko-electronic sting.

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.

Open Image to Video

1:1

Spin Dress Change — Living Room

Use this image as the first frame. Keep the exact same young Black woman — face, curly black hair, body proportions, pearl jewelry, and the living-room background (sofa, coffee table, kitchen doorway) locked to the reference. Photoreal image-to-video, continuous one take, no cuts. 0–0.8s: She stands centered, facing the camera, soft smile, both arms open slightly to the sides, champagne satin slip dress as in the reference, warm indoor light. 0.8–3s: First full spin left-to-right (clockwise from her left). As she turns through her back and comes around facing camera again, the dress morphs into a different elegant gown — new color and silhouette (e.g. deep emerald silk with soft draping), face and hair unchanged. 3–5.5s: Second continuous left-to-right spin. On the next face-front beat, wardrobe changes again to another elegant dress (e.g. midnight blue satin with a subtle flare), same jewelry, same room. 5.5–8s: Third left-to-right spin. Final face-front reveal into a third elegant look (e.g. soft blush chiffon floor-length), she finishes facing camera with an open-arm pose and a bright smile. Fabric swirls with realistic motion blur on turns; identity rock-solid. Three continuous spins, three elegant dress changes only at the face-front moments. No face morph, no extra people, no text, no logos, photoreal cinematic, 8 seconds.

3:4

Raincoat Neon — Cold to Wicked

Same young East Asian woman as the reference: wet black hair, clear plastic hooded raincoat with raindrops, pale dewy skin, cinematic rainy neon city night. Lock face, hairstyle, raincoat, and red-blue neon bokeh to the reference. Handheld medium shot, shallow depth of field, wet asphalt reflections. 0–1.2s: She stares into the camera with a cold, blank look — flat eyes, no smile, rain ticking on the plastic hood. 1.2–2s: Suddenly her expression flips into a wicked, seductive smirk, eyes sharp and playful. 2–3.5s: She turns away so her back faces the camera, then runs a few light steps forward away from camera into the neon street; raincoat flutters. 3.5–5s: She turns to face the camera again and walks backward toward the lens a few steps, eyes locked on camera. 5–6s: Both hands rise; she points both index fingers at the camera / screen, bright teasing smile. 6–6.8s: Short open laugh, one shoulder shake, still facing camera. 6.8–8s: She spins and sprints out of frame to the side; camera holds on rainy neon street. Natural continuous one-take motion, no cuts, no face morph, no extra people, photoreal cinematic, 8 seconds.

Text to Image

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

Open Text to Image

3:4

Raincoat Soft Neon

Beauty portrait in a translucent raincoat, soft neon bokeh, rain droplets as highlights, porcelain skin, romantic noir elegance, photoreal

AI-generated still life: hands holding a glowing tea bowl, warm side light over linen and wood textures

1:1

Tea Light Hands

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

Image to Image

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

Open Image to Image

Image-to-image result: the same face restyled into a soft-lit studio beauty portrait on a creamy backdrop

3:4

Selfie to Soft Studio Beauty

For Image1, keep exact likeness, transform into a soft-lit beauty studio portrait, creamy backdrop, natural skin texture, refined catchlights, high-end beauty campaign, no makeup plastic look

Image-to-image result: same pose and identity restyled into luxurious silk evening wear under editorial light

Photo to Silk Fashion Edit

Keep pose and identity. Restyle wardrobe into luxurious silk evening wear, elegant jewelry, fashion-editorial lighting, couture beauty finish, accurate face

3. Choosing a model

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

Video models

Kling, Seedance, and Veo cover different strengths in motion fidelity, speed, and look. Try the cheaper / faster tier first, then upgrade resolution or model when the composition is locked.

Image models

Nano Banana and GPT Image (and related options) suit beauty, product, and editorial stills. Higher quality tiers and more references cost more credits.

4. Learn from the showcase

Homepage showcase cards ship with real production prompts. Tap Recreate to drop a brief into Create, then edit wording, model, and aspect ratio.

Beauty still — Soft Beauty Headshot

Ultra refined beauty headshot, soft butterfly lighting, luminous skin, clean ivory backdrop, 85mm feel. Use Text to Image, then Image to Video if you want a subtle push-in.

3:4

Soft Beauty Headshot

Ultra refined beauty headshot of a young woman, soft butterfly lighting, luminous skin with natural pores, subtle catchlights, clean ivory backdrop, high-fashion beauty still, 85mm

Product still — Glass Serum Hero

Serum bottle with water droplets on frosted glass, blush-pink gradient, luxury advertising still. Strong I2V candidate: slow orbit or gentle push while locking the bottle design.

AI-generated skincare product shot: serum bottle on frosted glass with water droplets, blush-pink gradient backdrop

1:1

Glass Serum Hero

Premium skincare serum bottle with water droplets on frosted glass, soft gradient backdrop in blush pink, elegant reflections, luxury cosmetic advertising still, ultra sharp

Cinematic T2V — Neon Rooftop Chase

Photoreal chase language: subject identity, rainy neon night, handheld pursuit, one clear conflict beat. Start in Text to Video; keep “no dialogue / no subtitles” if you want a clean cut.

16:9

Neon Rooftop Chase

Photoreal live-action cyberpunk chase at rainy night, crisp high-resolution cinema detail, wet asphalt reflections, clean HDR contrast, one 35mm-equivalent lens and one uninterrupted controlled handheld pursuit shot. The courier is a lean East Asian man in his mid-20s, cropped black hair with a cyan underlight edge, thin scar along the jawline, matte black riding jacket, fingerless gloves, compact magnetic satchel locked to his chest. Keep one identical courier throughout. He flees exactly 4 distinct armored drone-riders; each has a unique chassis silhouette and melee tool; rider 4 is a heavy enforcer with a telescoping stun lance. Never reuse a rider. Every near-miss and strike is one-to-one. Show 4 separate contact events: sparks, shattered plating, wipeouts. No multi-target hits. Each next rider actively cuts him off, rams, or grapples. Rotate vault, slide-under, wall-run, and final lance parry without adjacent repetition. Every relocation: micro-anticipation; a hard cyan strobe fully hides him and he vanishes; camera keeps physical follow with city visible; a second local strobe reveals him already in the next escape or attack pose. No transit smear, no whip pan, no motion blur. Camera stays close with smooth arcs and pivots toward each strobe destination without snapping. No dialogue or subtitles. Night city ambience, tire hiss, 4 distinct impact accents, dry strobe cracks, accelerating synth hit.

I2V motion lock — Cliffside Linen

Reference-first: same outfit and coastal cliff, then timed beats (wind → walk → spin → smile). Classic Image to Video pattern — lock likeness, script the motion.

3:4

Cliffside Linen — Wind Spin

Same young woman as the reference: long wavy blonde hair, beige linen shirt and wide pants, gold jewelry, barefoot on a sunlit cliff above turquoise sea. Lock face, outfit, and coastal background to the reference. Handheld medium-wide shot, bright golden daylight, windy. 0–2s: She stands on the rock edge, hair and linen fabric whipping in strong wind, soft side glance. 2–4.5s: She turns toward camera, steps forward along the ridge, fabric flares, cinematic backlight. 4.5–6.5s: Arms open wide, spins once in place — linen swirls, hair arcs, sea glitter behind. 6.5–8s: Faces camera with a bright smile, wind still strong; camera slowly pushes in. Continuous one take, photoreal, no face morph, no extra people, 8 seconds.

Browse showcase on the homepage

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.

Aspect ratio

Pick framing the channel needs — 16:9 for YouTube-style, 9:16 for vertical, 1:1 for feeds. Match the ratio to the final crop so you do not waste motion outside the frame.

Duration

Shorter clips cost fewer credits and are easier to prompt well. Lock composition on a short take, then extend length once the look is right.

Free to start

New accounts get 10 credits on signup — enough for a first low-cost generation. No card required to try. See Pricing for packs and plans.

View pricing

6. Start creating

Open Create for the shared studio feed, or jump straight into a tool page. Recreate from showcase whenever you want a strong starting brief.

Open Create See Pricing

Guide FAQ

Do I need a different prompt style per model?

Use the same clear formula first. Then tweak per model only if needed — some prefer shorter beats, others tolerate richer camera notes. Compare outputs instead of rewriting from scratch every time.

Should I start from text or from an image?

If you already have a frame you love, use Image to Video or Image to Image. If the idea only exists as words, start Text to Video or Text to Image, then optionally animate the still.

How do I reuse a showcase prompt?

On the homepage showcase, tap Recreate. The brief lands in Create (or the current tool composer). Edit, switch model, set ratio, then generate.