AI video and movies, pay as you go

Go to an image generator like ChatGPT, or choose one below, and paste this.

Here's the example.

Split-screen character reference sheet, 2K, 9:16.

LEFT half: full-body standing, head-to-toe, facing camera, relaxed neutral stance.
RIGHT half: tight chest-up portrait of the SAME person, same wardrobe, same lighting.

Subject: 28, athletic, light freckles across the nose, short dark curls, wearing a navy hoodie and black joggers, quiet confidence.

Background: pure white seamless void, empty, no props.
Lighting: flat soft even studio light, frontal.
Skin: hyper-real — visible pores, fine lines, uneven tone, peach fuzz. No beauty filter.

Photorealistic. Identity must match exactly across both panels.

Here's the guide.

• Split-screen on flat white: left = full-body, right = chest-up
• Fill age, build, features, hair, wardrobe, demeanor
• Push real skin — pores, uneven tone, peach fuzz (no beauty filter)
• Once you have the sheet, never re-describe the face — only attach the image

Template to paste:

Split-screen character reference sheet, 2K, [16:9 or 9:16].

LEFT half: full-body standing, head-to-toe, facing camera, relaxed neutral stance.
RIGHT half: tight chest-up portrait of the SAME person, same wardrobe, same lighting.

Subject: [your character age], [build], [2–3 distinguishing features], [hair], wearing [wardrobe], [demeanor].

Background: pure white seamless void, empty, no props, no floor line distraction.
Lighting: flat soft even studio light, frontal, no rim, no colour cast, no dramatic shadows.
Skin: hyper-real — visible pores, fine lines, uneven tone, peach fuzz or light stubble, faint freckling. No beauty filter, no plastic skin, no glossy retouching.

Photorealistic. Identity must match exactly across both panels.

Go to an image generator like ChatGPT, or choose one below, and paste this.

Here's the example.

Empty boxing gym interior, 2K, 9:16 vertical.

Worn wooden floor with scuffs, heavy bags in the mid-ground, ropes of a ring visible on the right, chalk dust in the air.
Late-afternoon light from high windows on the left — warm shafts cutting through cooler shadow in the corners.
Materials: scuffed timber, worn leather bags, dusty air.
No people, no text, no logos. Photorealistic, natural colour, slight film grain.

Here's the guide.

• Empty place only — no character (separate plate from the sheet)
• Match the character sheet's aspect (9:16 or 16:9)
• Name light / weather and materials you can almost touch
• History paths: exact year + ban anachronisms; no people, text, or logos

Template to paste:

Empty [your location / place] interior, 2K, [9:16 or 16:9].

[2–3 concrete details of the space].
[Time of day / weather / light direction].
Materials: [surfaces you can almost touch].
No people, no text, no logos. Photorealistic, natural colour, slight film grain.

• Make a composite still first, or in video bind explicitly: @Image1 is the character… @Image2 is the location…
• Relight the person to match the place — drop the sheet's flat studio light.
• Feet planted, shadow direction matching the room, correct scale.
• Then a living hold: soft breathing, tiny weight shift — not a mannequin freeze.

Composite still:

Photoreal still. @Image1 is the person (face, hair, body, wardrobe — identity lock). @Image2 is the location (environment + lighting only).

Place @Image1 naturally inside @Image2 at [where in the frame]. Relight the person to match the location’s light direction and colour temperature — do not keep the white-studio light from the character sheet.
Feet grounded on the floor, contact shadow matching the room’s light. Correct scale for the space.
Match weather / colour temp: [time of day / weather from @Image2].
Same camera height as a documentary still. No beauty filter. No text, logos, or watermarks.

Living hold (short video):

@Image1 is the character (identity only). @Image2 is the location (environment + lighting).
@Image1 standing in @Image2 at [where in the frame]. Relight subject to match the room. Soft natural breathing, a natural blink, tiny weight shift. [ONE camera move or static hold]. Do not add subtitles. No logos.

Lock the face with stills first, count your angles, then embed naturally into an empty location plate. Prefer images + physical language over JSON.

JSON vs physical language

• JSON usually isn't required — it organizes your thinking. Models care about reference images + physical language (pores, blinks, feet planted), not curly braces.
• Structured blocks (REFERENCE MAP / TIMED BEATS / CONSTRAINTS) are fine.
• Wrapping the whole prompt in {json} rarely improves pixels. API JSON for settings (duration, aspect, seed) is separate.

JSON usually is NOT required for hyper-real stills or Seedance beats.
It organizes YOUR thinking — models care about reference images + physical language (pores, blinks, feet planted), not curly braces.
Structured text blocks (REFERENCE MAP / TIMED BEATS / CONSTRAINTS) are fine.
Wrapping the whole prompt in {json} rarely improves pixels.
API JSON for settings (duration, aspect, seed) is separate from the creative prompt.

Hyper-real consistent faces

• Lock stills first — split sheet = full-body + chest-up on flat white.
• If identity drifts, add 3/4 + profile panels of the same person (same wardrobe/hair/light).
• Practical Seedance recipe: 2–4 key refs (soft best 1–8). Bind @Image1 once — don't re-describe the face each beat.

Angle counts

• Minimum: 1 good sheet (split full-body + chest-up).
• Stronger: front + 3/4 + profile (~3–4) + body when you need orbit / walk-around lock.
• 8+ near-duplicates hurt — diversity beats volume.

Multi-angle lock sheet (paste):

Multi-angle face/body lock sheet, 2K, [9:16 or 16:9].

Panel A — front: chest-up, eyes to camera, neutral mouth, flat soft studio light on pure white.
Panel B — 3/4: same person, slight turn [left or right], same wardrobe + hair, same light.
Panel C — profile: same person, true side view, ear visible, same wardrobe + hair.
Optional Panel D — full-body front standing, head-to-toe, relaxed neutral stance (if not already on your split sheet).

Subject: [age], [build], [2–3 distinguishing features], [hair], wearing [wardrobe].
Skin: hyper-real — visible pores, fine lines, uneven tone, peach fuzz. No beauty filter.
Identity must match exactly across every panel. Pure white void background.

Natural embed

• Empty location as @Image2; place @Image1 in; relight to the room.
• Feet planted + contact shadow + correct scale.
• Living breath / weight shift — not a mannequin freeze.
• One camera move per beat; match weather/colour temp to @Image2.

Natural embed still (paste):

Photoreal composite. @Image1 is the person (face/hair/body/wardrobe — identity lock only). @Image2 is the empty location (environment + lighting only).

Place @Image1 naturally inside @Image2 at [where in the frame — e.g. mid-ground left of the doorway].
Relight the person to match @Image2 light direction and colour temperature — drop white-studio light from the sheet.
Feet planted on the floor with a soft contact shadow matching the room light. Correct scale for the space.
Match weather / colour temp to @Image2 ([time of day / weather]).
Living still energy: soft breathing, tiny weight shift, one natural blink — not a mannequin freeze.
Camera: [one framing — e.g. medium documentary still], eye-level. No beauty filter. No text, logos, or watermarks.

Seedance beat skeleton (paste):

REFERENCE MAP
@Image1 is the character (face, body, wardrobe — identity only; bind once — do not re-describe the face each beat).
@Image2 is the empty location (environment + lighting).

INVENTORY / CONTINUITY LOCKS
Same person as @Image1. Same place as @Image2. Relight to @Image2. Feet grounded + contact shadow + correct scale. Soft breathing, natural blinks, tiny weight shifts. Hair/wardrobe/weather locked.

TIMED BEATS (~8s)
0s-3s: [framing + ONE camera move]. @Image1 [subject action] in @Image2. Match weather/colour temp. [physical expression — not a mood word].
3s-6s: [framing + ONE different move]. [action that advances the plot]. Same light side.
6s-8s: [static hold or tiny push-in]. [payoff]. Hold final frame clean.

CONSTRAINTS
Do not add subtitles. No logos. Photoreal pores. One camera move per beat. Prefer Seedance. Practical refs: 2–4 images (soft best 1–8).

Add 2–4 images (practical Lucy recipe). In the prompt write a REFERENCE MAP, then timed beats with Seedance syntax 0s-3s / 3s-7s.

• Seedance 2.0: up to ~9 refs · Seedance 2.5: up to ~30 images (soft best with 1–8 image subjects)
• Typical hyper-real set: @Image1 character sheet · @Image2 location · optional @Image3 product · optional @Image4 start keyframe
• Bind every image: “@Image1 is the character (face, body, wardrobe — identity only, not lighting/background)”
• Each beat: ONE camera move + subject action + light continuity + physical expression + optional dialogue/SFX — enough plot, not overloaded

Full paste template:

REFERENCE MAP
@Image1 is the character (face, body, wardrobe — identity only, not lighting/background).
@Image2 is the location (environment + lighting).
@Image3 is [your product — optional label/shape lock].
@Image4 is [start keyframe — optional opening composition].

How many images: typically 2–4 for hyper-real Seedance clips (direct Seedance; not hyper-real people through Lucy). Seedance 2.0 accepts up to ~9 refs; Seedance 2.5 up to ~30 images (soft best with 1–8 image subjects). Never re-describe the face once the sheet exists — attach @Image1.

INVENTORY / CONTINUITY LOCKS
Same person as @Image1. Same place as @Image2. Relight subject to @Image2 (drop white-studio light). Feet grounded, contact shadow matching room light, correct scale. Soft breathing, natural blinks, tiny weight shifts. Hair / wardrobe / weather locked. Photoreal pores — no beauty filter.

TIMED BEATS (~8s — Seedance 2.0/2.5 friendly)
0s-3s: [framing + ONE camera move]. @Image1 [subject action] in @Image2. Light/weather continuity. [physical expression — not a mood word]. Soft ambience.
3s-6s: [framing + ONE different move]. [subject action that advances the plot]. Same light side. Optional {[short dialogue]} or <[sfx]>.
6s-8s: [framing + static hold or tiny push-in]. [payoff action]. Hold final frame clean.

CONSTRAINTS
Do not add subtitles. No logos/watermarks. Photoreal pores. Relight to scene. One move per beat. Matching shadows/scale. Prefer Seedance for this hyper-real path (Veo/Kling ok as alternatives).

Pick a path, add the noted images, paste the prompt into Seedance (or try Veo/Kling). Each block includes image map, timed beats, and why the camera moves were chosen.

REFERENCE MAP
@Image1 is the character (face, body, wardrobe — identity only, not lighting/background).
@Image2 is the location (environment + lighting).

INVENTORY / CONTINUITY LOCKS
Same person as @Image1. Same place as @Image2. Relight subject to match @Image2 light direction and colour temperature. Soft breathing + tiny weight shifts. Hair, wardrobe, weather locked. No beauty filter.

TIMED BEATS (~8s)
0s-3s: Cinematic wide, slow gentle tracking beside her (choose tracking to keep her moving through space without a cut). @Image1 walks slowly through @Image2, dust or light haze drifting, feet planted with correct scale and contact shadow. Dramatic side light, photoreal 4K. Soft ambient bed.
3s-6s: Medium shot, slow push-in (choose push-in when the emotion tightens). She stops; shoulders drop on a long exhale; jaw softens; gaze holds past the lens — no smile yet. One blink. Continuity: same light side as beat 1.
6s-8s: Close-up, static hold (choose static to land the beat clean). Side light rakes pores and fine lines; hair edge drifts once; hold the final frame clean.

CONSTRAINTS
Do not add subtitles. No logos/watermarks. Photoreal pores. Matching shadows/scale to @Image2. One camera move per beat. Prefer Seedance.

REFERENCE MAP
@Image1 is the character (face, body, wardrobe — identity only, not lighting/background).
@Image2 is the location (bright bathroom or desk — environment + lighting).
@Image3 is the product (label, shape, colour — packaging lock).

INVENTORY / CONTINUITY LOCKS
Same person as @Image1. Same room as @Image2. Same product as @Image3. Relight subject to morning/afternoon window light from @Image2. Soft breathing throughout. Product shade/label never changes. No beauty filter.

TIMED BEATS (~8s)
0s-2s: Medium selfie-style handheld sway, eye-level (choose handheld for UGC authenticity). @Image1 holds @Image3 toward camera in @Image2, twists the cap or turns the pack so the label reads. Soft room tone, no music.
2s-5s: Close-up, slight handheld sway. She uses @Image3 in one clear demo action, presses lips or nods once, corner of mouth tugs into a real small smile. {"Okay, this actually works."}
5s-8s: Medium close-up, static hold (choose static so packaging stays sharp). She turns toward window light, holds @Image3 beside her face; pores and peach fuzz visible; silent beat — small satisfied exhale.

CONSTRAINTS
Do not add subtitles. No logos/watermarks beyond the product label. Photoreal pores, iPhone UGC look. One camera move per beat. Relight to @Image2.

REFERENCE MAP
@Image1 is the character (face, body, wardrobe — identity only). Wardrobe stays modern Gen Z streetwear forever — never period costume.
@Image2 is the location (period environment + lighting only — exact year locked). Empty of modern people/props in the still.

INVENTORY / CONTINUITY LOCKS
Host identity = @Image1 only. Era = @Image2 only (e.g. London street, 1666 — timber frames, cobbles, smoke haze, late-afternoon amber light). Locals in background wear period dress only; host never changes into period clothes. Relight host to @Image2. Soft breathing, natural blinks, tiny weight shifts while walking. Ban anachronisms: no cars, phones, plastic, LED signs, modern logos, sneakers on extras.

TIMED BEATS (~8s)
0s-3s: Medium handheld selfie / walk-and-talk, slight sway (choose handheld so it feels like a phone vlog). @Image1 enters @Image2 from frame left, walking toward camera then turning to show the street; modern outfit contrasted against timber and smoke. {"Okay — so this is London, 1666, right before everything burns."} Distant period crowd murmur, wood-cart wheels, no music.
3s-6s: Medium, slow tracking beside her (choose tracking to keep walking without a cut). She watches a smoke plume rise over rooftops; one slow blink; shoulders tense on a sharp inhale; points once with her free hand. Soft wind, distant shout in period language (do not subtitle).
6s-8s: Medium, handheld as she walks away then glances back over her shoulder (choose walk-away for a closing thought). {"And nobody here knows what's coming."} Hold final frame clean — pores readable, no beauty filter.

CONSTRAINTS
Do not add subtitles. No logos/watermarks. Photoreal. Host wardrobe locked modern. Period accuracy locked to @Image2. One camera move per beat. Prefer Seedance.

Path tip — open Camera + expression craft and Lessons from great directors below for animated move/expression/technique previews and Seedance-ready Copy prompt phrases (one move per beat).

More product-ad results: Our review of the AI models · /product-showcase/veo_generic_cup.mp4 and Harper clips under /product-showcase/.

Animated previews + Seedance-ready Copy prompt phrases for camera moves and physical expressions, then mix-and-match craft chips (skin, lighting, @Image counts, beats).

More craft chips below — open one, copy a line into your Seedance prompt. The Camera angles chip is the text summary; prefer the visual choosers above.

Paste into stills or CONSTRAINTS:
Skin: hyper-real — visible pores, fine lines, uneven tone, peach fuzz or light stubble, faint freckling. No beauty filter, no plastic skin, no glossy retouching.

Prefer the visual Expression chooser above (animated cards + Copy prompt) — physical cues Seedance follows, never mood words alone.
Quick reminder:
• Shoulders drop; a long exhale; jaw unclenches.
• One eyebrow lifts; corner of the mouth tugs, then settles.
• She soft-blinks once then holds gaze at the lens.
• Lips press together, then break into a small real smile.
• Tiny weight shift foot-to-foot; fingers fidget once on the product.
• Chin tips up; nostrils flare on a sharp inhale before she speaks.

Empty place still first — no character.
Name: [place] + [exact year if history] + [2–3 materials] + [light/weather].
Crowds: only if period-safe; ban cars, phones, plastic, LED, modern logos.
Match aspect to the character sheet ([9:16 or 16:9]).

Prefer the visual Camera move chooser above (animated cards + Copy prompt) — one move per beat; phrases describe what the frame does so Seedance follows.
Quick reminder:
• Static lock-off — land a payoff / read a label.
• Dolly in / push-in — intimacy when emotion tightens.
• Dolly out / pull-back — reveal scale after a tight beat.
• Track left/right — walk-and-talk without cutting.
• Pan / tilt — motivated look inside one space.
• Handheld sway — UGC / selfie energy (not shake-cam).
• Orbit / arc — spectacle or product hero.
• Crane / rise — open on a world or lift out.
• Whip pan — fast blur between two clear end-frames (sparingly).
• Rack focus — redirect attention without moving the camera.
Snippet: Camera: medium shot, slow push-in, eye-level. One move only. Hold the final frame clean.

On every living hold or quiet beat:
Soft natural breathing, a natural blink every few seconds, tiny weight shift foot-to-foot. Hair or coat edge may drift. Not a mannequin freeze.

Say what moves vs what stays:
Background: light haze / dust drifting; distant crowd soft blur; wind in banners.
Subject: mostly locked identity — only micro-breathing and blinks.
Useful for bullet-time, history streets, and cinematic wides.

Relight the person to match the location — drop the character sheet's flat studio light.
Match contact-shadow direction and colour temperature to the room.
Continuity: hair, wardrobe, product label, weather locked across beats.
Always: Do not add subtitles. No logos/watermarks unless you want them.

Add N images, then bind them in the prompt:
• Seedance 2.0: up to ~9 refs. Seedance 2.5: up to ~30 images (soft best results with 1–8 image subjects).
• Practical Seedance hyper-real recipe (best connecting to Seedance directly — Lucy offers Seedance but cannot get hyper-real people results through Lucy): typically 2–4 images — e.g. @Image1 character sheet (or chest-up), @Image2 location, optional @Image3 product / @Image4 start keyframe.
• Write: "@Image1 is the character (face, body, wardrobe — identity only, not lighting/background)."
• "@Image2 is the location (environment + lighting)."
• "@Image3 is the product" if needed.
• Use the same labels in every timed beat. Never re-describe the face in text once the sheet exists — attach the image.

Seedance timestamp syntax: 0s-3s / 3s-7s (NOT 0:00-0:03).
Each beat: ONE camera move + subject action + light/weather continuity + physical expression (not mood words) + optional dialogue/SFX.
Pack enough plot that the model doesn't improvise freely; don't overload or you'll get frantic cuts.
Friendly lengths: ~8s clips for Seedance 2.0 / 2.5.

[CHARACTER]
A synthetic pursuer built for one purpose: relentless, unblinking pursuit. Broad-shouldered chrome endoskeleton visible through tears in scorched synthetic skin along one forearm, glowing red optical sensors, torn leather jacket, combat boots. Moves with mechanical, unnervingly steady precision - no hesitation, no fatigue.

[SCENE]
A derelict multi-story parking garage at night. Flickering fluorescent tubes, concrete pillars streaked with rust, oil pooling under abandoned cars, a single exit ramp spiraling down into darkness.

[SHOT SEQUENCE]
SHOT 1 (0s-2s): Low-angle tracking shot, camera mounted street-level beside the motorcycle. The pursuer guns the engine, front wheel lifting slightly, sparks skittering off a support pillar as the handlebar clips it.
SHOT 2 (2s-5s): Handheld chase cam, whip-panning between the bike and a support column - a fuel-drum rupture kicks an orange fireball skyward, trailing black smoke, debris scattering across the oil-stained floor.
SHOT 3 (5s-7s): Close-up, static camera. The pursuer's face lit red by the fireball's glow - no fear, no flinch, optical sensors narrowing with mechanical focus.
SHOT 4 (7s-8s): Wide shot, camera holds as the bike bursts through the exit-ramp shutter in a shower of sparks and torn metal, disappearing into the night.

[CONSTRAINTS]
No subtitles/logos/watermarks. Cinematic, high-contrast, 4K, desaturated blue-grey palette except the fire's orange glow.

[REFERENCE]
@Image1 - your character reference (from Cast & Locations). Use for face, hair, skin tone, and build only - not background or lighting. @Image2 - the lipstick.

[CHARACTER]
Mid-20s, warm brown skin, natural curls pulled into a loose bun, silky blush-pink slip dress, relaxed and confident.

[SCENE]
A bright, clean bathroom. Morning light through a frosted window, softly catching the fabric of her dress and the edge of the mirror.

[SHOT SEQUENCE]
SHOT 1 (0s-2s): Medium shot, static camera, mirror reflection. She twists open @Image2, inspecting the shade with a small approving nod.
SHOT 2 (2s-5s): Close-up, slight handheld sway (selfie-style). She applies it in one smooth stroke, presses her lips together, breaks into a genuine, pleased smile. {"Okay, this shade is unreal."}
SHOT 3 (5s-8s): Medium close-up, camera holds. She turns toward the window light, holding the product beside her face so the label reads clearly, natural light catching both skin and packaging.

[CONSTRAINTS]
No subtitles/logos beyond the product's own label. Warm, soft-focus, natural light, iPhone-shot UGC aesthetic - not overly polished.

[REFERENCE]
@Image1 - your character reference. @Image2 - the product.

[CHARACTER]
Late 20s, short textured hair, glasses, oversized knit sweater, easygoing and a little wry.

[SCENE]
A cozy bedroom desk setup, string lights soft in the background, laptop open, afternoon light through a nearby window.

[SHOT SEQUENCE]
SHOT 1 (0s-3s): Medium shot, static camera, desk-level. He picks up @Image2, turning it over in his hands, one eyebrow raised, genuinely impressed. {"Okay, I was not expecting this to actually be good."}
SHOT 2 (3s-6s): Close-up, slow handheld push-in. He demonstrates the product's main feature to camera, focused and matter-of-fact.
SHOT 3 (6s-8s): Medium shot, camera holds. He sets it down, leans back, shrugs with a small grin. {"Yeah. It's going on the desk permanently."}

[CONSTRAINTS]
No subtitles/logos beyond the product's own branding. Casual, natural light, handheld UGC energy - not a polished commercial.

REFERENCE
@Image1 — [who/what]. Use for face/body/wardrobe/identity only, not background or lighting.
@Image2 — [empty location / product — optional].

CHARACTER (only if you still need text — prefer attaching the sheet)
[age] + [build] + [distinguishing features] + [hair] + [wardrobe] + [demeanor]

SCENE
[where], [when], [atmosphere], [lighting/color tone] — 2–3 sentences

SHOT SEQUENCE
SHOT 1 (0s-3s): [framing + ONE movement] — [subject action]. {[short dialogue]}
SHOT 2 (3s-6s): [framing + movement] — [subject action]. ([music note])
SHOT 3 (6s-8s): [framing + movement] — [subject action]. <[sfx note]>

CONSTRAINTS
Do not add subtitles. No logos/watermarks unless wanted + [style anchor]

Director technique pack — cinematic / director lessons, classic feature grammar, ads / hero-reveal, and music-video camera language as filterable chooser cards. Each card has an original CSS-animated SVG mini-loop, a one-line lesson, when to use it on Lucy, and a Seedance-ready Copy prompt. No trailers, ad rips, or music-video footage.

Break the story into beats. Generate each beat as its own short clip, then combine in /stitch.

Draft cheap first, then upscale the keeper. Sound: write per-shot dialogue/SFX, or say "no music" and add score later in /stitch.