50 AI Video Prompts That Actually Work (Tested on WAN 2.2, Sora, Runway, Kling)
Copy-paste library of 50 image-to-video prompts across 5 categories — anime, portraits, cinematic, product, NSFW. Each prompt tested and tagged with the model it works best on, plus model-specific writing tips.
AI video prompts are different from text-to-image prompts. With an image-to-video model, the source image already carries the scene — the prompt only needs to describe what should change between frames. Most beginners over-prompt (re-describing the scene), under-prompt (giving the model nothing to anchor on), or fight the model’s preferred vocabulary.
This guide is a 50-prompt library, organized by category, with each entry copy-pasteable. Every prompt has been tested on at least one of WAN 2.2, Sora 2, Runway Gen-3, or Kling 1.6, with the model noted on each card.
How to write an AI video prompt — 5 rules
1. Subject + action + camera + style
Every effective prompt has four components in roughly this order: what subject moves, what action they perform, how the camera behaves, and what style governs the look.
2. Specify motion explicitly
I2V models default to over-animating when you under-specify. Always name the body part or object that moves, the direction, and the speed. “Hair sways” is weak; “hair sways gently from left to right over 3 seconds” is strong.
3. Reference real cinematography
Models trained on film footage respond to film vocabulary. Use terms like dolly push-in, crane down, handheld follow, shallow depth of field, slow-motion, cinematic framing.
4. Use anchor words by model
WAN 2.2 likes slowly, subtle, gentle. Kling likes cinematography vocabulary. Runway responds to film-references. Hailuo is most prompt-faithful. See the model-specific tips section below.
5. Iterate, don’t perfect
First generation rarely lands. Plan 3-5 attempts per concept, and change one word at a time when iterating — easier to find which word made the difference.
10 anime animation prompts
anime girl, gentle wind blowing hair from left to right, hair strands lifting slowly, subtle smile forming, soft breathing chest motion, eyes blink every 4 seconds, camera completely static, atmospheric lighting unchanged
anime character with flowing magical cape, cape swirls and lifts dramatically in slow motion, character holds confident pose, magical particles drift around, single 5-second action loop, dramatic lighting unchanged
anime warrior slowly draws sword from sheath in slow motion, blade catches light at peak, single smooth motion over 4 seconds, character expression intense, no camera move
anime mage raises both arms slowly, magical energy gathers between palms with growing glow, hair lifts from energy buildup, soft light pulse, motion peaks at the end of clip, no camera move
anime fighter shifts from idle to combat stance over 2 seconds, hair flows with motion, hands raise to guard, eyes narrow with focus, dramatic action lines appear briefly, camera static
anime girl, eyes slowly fill with glistening tears, single tear runs down cheek, lip slightly trembles, hair gentle sway, sad atmospheric lighting unchanged, no camera move, single 4-second motion
anime character in static pose, soft natural breathing motion, eyes blink every 3 seconds, hair gentle sway, single seamless 5-second loop, no camera move, lighting unchanged — perfect for VTuber-style idle
anime girl performs slow dramatic hair flip from one side to the other, hair sweeps across in slow motion, face follows the motion with confident smirk, single 4-second motion, camera static, lighting unchanged
anime character begins powering up, energy aura pulses around body, hair lifts from intensity, ground particles float upward, single energy build over 5 seconds peaking at end, dramatic lighting deepens
anime girl slowly raises eyes toward camera, head tilts up slightly, soft confident smile develops, hair gentle sway, single smooth motion over 3 seconds, no camera move, lighting unchanged
10 portrait / character prompts
subject in static cinematic portrait pose, soft natural breathing, single slow blink mid-clip, slight head tilt at end, depth of field shallow, camera completely static, lighting unchanged, magazine cover aesthetic
subject starts neutral, soft smile slowly develops over 3 seconds reaching wide warm smile, eyes brighten, gentle hair sway, no camera move, lighting and background unchanged
subject slowly turns head 30 degrees to the right then back to center, single smooth motion over 4 seconds, eyes follow, soft confident expression, no camera move, lighting unchanged
subject slowly raises eyes from looking down to direct eye contact with camera, soft confident smile develops, slight head tilt, single smooth motion over 3 seconds, no camera move, atmospheric lighting unchanged
subject's expression shifts from neutral to surprise — eyes widen, mouth opens slightly, eyebrows raise, single rapid motion over 1 second then holds, slight backward head tilt, no camera move
subject slowly bites lower lip, eyes look up at viewer through lashes, single smooth sensual motion over 2 seconds, hair gentle sway, soft natural light, no camera move
subject slowly raises hand to chin, fingers rest thoughtfully against jawline, eyes look slightly up and to the side, contemplative expression, single smooth motion over 3 seconds, no camera move
subject performs slow flirty wink with right eye, follows with confident smile, slight head tilt, single seamless 2-second motion, hair gentle sway, no camera move, lighting unchanged
subject slowly raises hand to face, tucks loose strand of hair behind ear, soft smile develops, eyes maintain contact with camera, single smooth motion over 4 seconds, no camera move
subject slowly looks back over right shoulder toward camera, hair sweeps with motion, sultry confident expression at peak of turn, single smooth motion over 3 seconds, no camera move
10 cinematic / scene prompts
slow cinematic dolly push-in from medium shot to medium close-up over 5 seconds, subject still with breathing motion, depth of field shallow throughout, soft warm light unchanged
slow cinematic crane shot descends from above subject's shoulder height to chest level, revealing setting around them, subject still with breathing motion, light wraps as camera descends, depth of field shallow
static scene, soft particles (dust, snow, embers) drifting from upper-right to lower-left across frame, atmospheric haze drifts slowly, gentle warm light flicker, single 5-second seamless loop, camera completely static
handheld camera follows subject walking forward, subtle camera shake matches walking rhythm, subject in soft focus, background drifts slightly, cinematic warm grade, 5-second motion
drone shot starts top-down on subject, slowly tilts forward and pulls back to reveal wide environmental shot, single smooth motion over 5 seconds, subject still, cinematic landscape lighting
rack focus pull from foreground subject to background detail over 4 seconds, subject in foreground holds still pose, blur shifts smoothly, depth of field shallow, no camera move
static wide shot of landscape, gentle wind moves grass and tree leaves from left to right, distant clouds drift slowly across sky, atmospheric soft light shifts subtly, single 5-second loop
static scene composition, lighting slowly transitions from late afternoon golden hour to dusk, shadows lengthen across frame, color temperature warms, subject still with breathing motion, single 5-second motion
static outdoor scene, rain begins to fall — first sparse drops, building to gentle steady rainfall by clip end, subject still, atmospheric haze increases, lighting cools slightly, single 5-second motion
slow cinematic zoom out from close-up to wide shot over 5 seconds, subject in center remains still, surrounding context revealed gradually, depth of field deepens as zoom progresses
10 product / marketing prompts
product rotates slowly 360 degrees, single smooth rotation over 5 seconds, lighting wraps consistently around it as it turns, soft floor reflection unchanged, no camera move, studio backdrop static
static product on dark background, dramatic light sweep moves across product from left to right over 4 seconds, revealing details and textures, subtle highlight on edges, no camera move
slow-motion liquid pour from bottle into glass, droplets form and arc through air, splash rises gracefully on impact, single 5-second motion, soft studio lighting, depth of field shallow, no camera move
apparel item (dress / scarf) catches gentle wind, fabric flows and ripples in slow motion, single smooth motion over 4 seconds, soft natural lighting, depth of field shallow, mannequin / model still
static plated dish, steam rises softly from surface in delicate curls, single seamless 5-second loop, warm restaurant lighting unchanged, depth of field shallow, no camera move
tech device on clean background, UI elements on screen pulse and animate subtly, soft glow around device, single 5-second loop, studio lighting unchanged, no camera move
cosmetic product (lipstick / nail polish) drops single droplet onto white surface, droplet impacts and spreads in slow motion, splash forms petal pattern, single 5-second motion, soft studio lighting
static shoe on white background, slow camera tilt from low angle to slightly elevated angle, revealing details of design and construction, single smooth motion over 4 seconds, soft studio lighting
luxury watch face fills frame, second hand sweeps slowly through one minute over 5 seconds, subtle glint catches light, no camera move, soft studio lighting unchanged
brand logo on dark background, particles assemble inward to form logo, soft glow pulse on completion, single 4-second motion, atmospheric haze drifts slowly, no camera move
10 NSFW-safe (anime / fantasy only) prompts
subject lying back, gentle rhythmic rocking motion of body, hair flows with motion, soft confident expression with eyes half-closed, single seamless 4-second loop, soft warm bedroom lighting unchanged
subject slowly slides garment strap off shoulder, fabric slides down revealing skin, sultry eye contact with camera, single smooth motion over 5 seconds, hair sways with motion, soft natural light unchanged
subject slowly slides own hand from collarbone down across body, sultry confident expression, eyes half-closed in pleasure, slow sensual motion over 5 seconds, hair gentle sway, no camera move
subject riding in slow cowgirl rhythm, body rises and falls with each motion, hair sways with rhythm, lips parted in pleasure, single seamless loop over 4 seconds, soft warm bedroom lighting unchanged
subject in shibari rope harness, gentle struggle motion, chest rises and falls with deep breathing, ropes pressing slightly into skin as she shifts, lips parted, single 4-second loop, dim moody lighting unchanged
subject's expression slowly transitions from neutral to ahegao — eyes roll back, tongue extends out, blissful intense pleasure expression develops over 3 seconds, slight head tilt back, soft moaning lips, no camera move
subject's body arches gradually as orgasm builds, head tips back, mouth opens, hands grip sheets, intense climax peak with arch at peak then slow relaxation, single 5-second motion, lighting unchanged
subject lying back, legs slowly spread open from closed to wide, hand resting on inner thigh, sultry confident expression, slow seductive motion over 5 seconds, soft bedroom lighting unchanged, no camera move
bare breasts gently bounce in subtle continuous rhythm, soft natural sway, nipples slightly hardening, subject's body otherwise still with soft breathing, single seamless 3-second loop, soft warm lighting unchanged
subject lying back with one hand between own legs, slow rhythmic fingering motion, other hand grabs own breast, head tipped back, mouth slightly open, eyes closed in pleasure, single 4-second loopable cycle, soft bedroom lighting
Model-specific tips
WAN 2.2 — what works
Verbs like slowly, subtle, gentle are critical — WAN 2.2 over-animates without them. NSFW content works out of the box (no built-in filter on the open-weights model). Strong on stylized / anime sources. Weaker on dramatic camera moves than Kling or Hailuo.
Sora — what works
Most coherent at long generation lengths (10+ seconds without identity drift). Photoreal input handles best. Strict NSFW filter. Best for cinematic and complex scenes.
Runway Gen-3 — what works
Responds to film-references and DP names (“Roger Deakins lighting,” “Wes Anderson framing”). Pro tooling (motion brush, multi-reference). Highest ceiling, highest cost. NSFW filtered.
Kling — what works
Best identity preservation in this generation of models. Responds to camera vocabulary (dolly, crane, push-in). Aggressive NSFW filter. Strong on people-centric content.
Hailuo — what works
Most prompt-faithful of the four. What you describe is what you get. Strong on cinematic camera moves. NSFW filtered. Generous free tier (1000 credits/mo).
How to remix prompts
Pick any prompt above, change one of these to spawn variants:
- The body part / object that moves— “hair sways” → “dress flows” → “cloak billows”
- The direction— “left to right” → “forward to back” → “rising upward”
- The speed— “gentle” → “dramatic” → “in slow motion”
- The camera— “static” → “slow push-in” → “handheld follow”
- The lighting— “warm sunset” → “cool moonlight” → “dim moody dungeon”
Five variables × five values each = 3,125 distinct prompts from the 50 above. Don’t over-think it — most variants land in the first 1-3 attempts on a strong base prompt.
Frequently asked questions
Which model takes which kind of prompt?
See the model-specific tips above. Short version: WAN 2.2 wants slow / subtle verbs; Kling wants camera vocabulary; Runway wants film references; Hailuo wants exactly what you describe.
How long should an AI video prompt be?
1-3 sentences max. The source image already carries the scene; the prompt is purely about motion. Over-prompting (re-describing what’s already in the image) usually hurts quality.
Can I use these prompts commercially?
Yes — prompts aren’t copyrightable. The output (video) is a separate question and depends on the tool’s license: most paid tiers grant commercial rights; most free tiers don’t.
Why does my prompt produce different output every time?
I2V models are stochastic — same prompt + different random seed = different output. Lock the seed if you want reproducibility, or just generate 3-5 attempts and pick the best.
Where can I test these prompts for free?
For NSFW prompts, DFP — the only free I2V tool that permits explicit content. For SFW, Hailuo’s 1000 free monthly credits cover ~20 attempts.
Try it yourself — free
Every new account gets 20 free credits. Run an edit, a faceswap or an animation in under a minute.
