[Guide] How We Made “She Dropped Her Bag” Video with GenX Pro App

The story

Warm afternoon light down a long avenue. Yellow taxis, stone buildings, a glossy black sedan at the curb.

She gets out. He's coming the other way with the dog. Her bag falls. Two hands reach for it at the same time, they look up, and they hold the look.

That's it. No line, no voiceover, no music until the end.

What's The Generation Sequence for This?

Steps 1–3 — The cast

Three reference sheets, same recipe each time: one subject, three panels, grey studio background. A front wardrobe view, a back view, and a tight hero portrait.

Why three panels and not one portrait. The right panel locks the face. The left and centre lock the clothes and the proportions, from every side. Give the video model only a face and it will invent the back of the coat — differently in each shot.

The headless front panel is deliberate. It's a garment display, so the model reads it as wardrobe rather than as a second person.

Step 1 — The woman

MODEL: GenX Pro -> Image → Nano Banana Pro.

Settings: 16:9.

Create a photorealistic 16:9, 1920 x 1080 production character sheet arranged
as exactly three equal vertical panels separated by thin white lines, all on one
uniform neutral light-gray seamless studio background with soft clean catalog
lighting. Show the same original woman consistently across all panels. She is
29, light-olive skinned, with an oval face, hazel-green eyes, softly arched
brows, refined straight nose, natural lips, a subtle beauty mark below her left
cheekbone, and shoulder-length chestnut-brown hair in soft late-1980s waves with
a side part. Locked outfit: deep burgundy belted trench coat over a cream silk
blouse, dark knee-length A-line skirt, sheer neutral tights, dark-brown
low-heeled pumps, small round gold earrings, slim gold wristwatch, and
structured dark-brown leather shoulder bag. LEFT: complete straight-on front
wardrobe view from a clean neck base through both shoes; head intentionally
absent as a non-graphic editorial garment display, with visible margin above the
neck and below the shoes so the body is not cropped; both hands visible; bag on
left shoulder. CENTER: complete back view from head to shoes, showing hairstyle,
trench back, belt, bag strap, skirt, legs, and heels. RIGHT: tight front-facing
head-and-shoulders hero portrait of this same woman, calm neutral expression,
highly detailed natural face and hair. Keep identity, outfit, proportions, skin
tone, and lighting identical across panels. Avoid men, neck artifacts, celebrity
resemblance, inconsistent faces or clothing, extra limbs or fingers, malformed
hands, warped shoes, duplicate accessories, readable text, logos, brands, modern
objects, dramatic posing, heavy makeup, plastic skin, blur, colored backgrounds,
extra panels, labels, captions, and watermark.

The burgundy trench is the continuity anchor. One saturated colour nothing else in frame is wearing — checkable at any distance, in any crop. Small props can't do this job.

Name the specifics. A beauty mark below the left cheekbone, a slim gold wristwatch. Concrete details give the model something to hold onto across cuts; "attractive woman, thirties" gives it nothing.

Step 2 — The man

MODEL: GenX Pro -> Image → Nano Banana Pro.

Settings: 16:9.

Create a photorealistic 16:9, 1920 x 1080 production character sheet arranged
as exactly three equal vertical panels separated by thin white lines, all on one
uniform neutral light-gray seamless studio background with soft clean catalog
lighting. Show the same original man consistently across all panels. He is 31,
with warm medium skin, deep-brown eyes, strong straight brows, a slightly
asymmetrical nose, defined but approachable features, a faint left-cheek dimple,
clean-shaven jaw, and short dark-brown wavy hair with natural volume. He must
not resemble a celebrity. Locked late-1980s outfit: navy suede bomber jacket
over a pale-blue Oxford shirt with open collar, charcoal pleated trousers,
dark-brown leather belt, clean off-white leather high-top sneakers, and a simple
steel wristwatch. LEFT: complete straight-on front wardrobe view from a clean
neck base through both shoes; head intentionally absent as a non-graphic
editorial garment display, with margin above the neck and below the shoes so the
body is not cropped; arms relaxed and both hands visible. CENTER: complete back
view from head to shoes, clearly showing hairstyle, jacket construction,
trousers, and sneakers. RIGHT: tight front-facing head-and-shoulders hero
portrait of this same man with calm neutral expression and highly detailed
natural face and hair. Keep identity, outfit, proportions, skin tone, and
lighting identical across panels. Avoid bandanas, neck artifacts, facial hair,
sunglasses, hats, celebrity resemblance, inconsistent faces or clothing, extra
limbs or fingers, malformed hands, warped shoes, logos, readable text, modern
streetwear, dramatic posing, plastic skin, blur, colored backgrounds, extra
panels, labels, captions, and watermark.

"He must not resemble a celebrity" earns its place. Ask for a handsome man in a bomber jacket in 1989 and you will drift toward someone recognisable.

Step 3 — The dog

MODEL: GenX Pro -> Image → Nano Banana Pro.

Settings: 16:9.

Create a photorealistic 16:9, 1920 x 1080 animal reference sheet arranged as
exactly three equal vertical panels separated by thin white lines, all on one
uniform neutral light-gray seamless studio background with soft clean catalog
lighting. Show exactly the same small wire-haired fox terrier consistently
across all panels. The dog has a mostly white wiry coat, tan-and-black markings
around the ears and eyes, dark-brown eyes, a black nose, naturally folded
V-shaped ears, a compact athletic body, proportionate legs and paws, and one
red-brown leather collar with a small plain brass ring and no readable tag.
LEFT: complete front standing view from ears to paws, entire dog visible with
margin on every side. CENTER: complete rear three-quarter standing view showing
the same coat markings, tail, body proportions, collar, and all four paws.
RIGHT: tight front-facing head-and-shoulders hero portrait with alert but gentle
expression and detailed wiry fur. Keep markings, anatomy, scale, collar, and
lighting identical across panels. Avoid extra dogs, changed markings, extra legs
or paws, fused limbs, deformed muzzle, duplicated ears, aggressive expression,
open mouth, leash, clothing, costumes, logos, readable text, colored
backgrounds, extra panels, labels, captions, and watermark.

Animals need a sheet as much as people do. Coat markings drift between shots worse than faces, and a dog that changes pattern mid-film is the one thing every viewer notices.

Leave the leash off the sheet. It belongs to the video prompt, where it has to move.

Step 4 — The street

MODEL: GenX Pro -> Image → Nano Banana Pro.

Settings: 16:9.

Create a photorealistic 16:9 wide environment plate of a bright, busy upscale
metropolitan avenue in the late 1980s, visually inspired by classic New York
fashion-commercial street photography. Position the camera at pedestrian eye
level on a broad sidewalk, looking diagonally down the avenue with a natural
35mm documentary perspective and a large, clean foreground staging area where
principal characters can be added later. Place a glossy black boxy late-1980s
luxury sedan at the curb in the near-middle ground, with additional
period-appropriate sedans and yellow taxis farther down the avenue. Every
vehicle must retain a physically realistic license-plate mount, but every plate
surface must be completely blank and unnumbered, with no letters, digits,
symbols, state names, or logos. Include tall stone and masonry commercial
buildings, storefront windows, modest fabric awnings, traffic lights, street
poles, curb details, and layered city depth. Add only a few naturally spaced
pedestrians in the middle and far background wearing understated late-1980s
clothing; keep the immediate foreground empty. Use deep focus similar to a real
street photograph captured around f/8: foreground pavement, parked cars,
pedestrians, storefronts, and distant architecture should remain clearly
recognizable and naturally detailed, with only subtle atmospheric softening in
the extreme distance. Use strong warm afternoon sunlight, realistic pavement
shadows, slightly muted warm colors, authentic 35mm film color and fine natural
grain, restrained highlight bloom, and polished but believable fashion-ad
realism. Avoid shallow depth of field, portrait mode, artificial bokeh, creamy
blurred backgrounds, extreme lens blur, miniature or tilt-shift appearance,
principal man, principal woman, dog, foreground person, staged interaction,
readable shop names, recognizable brands, visible license-plate characters,
modern vehicles, smartphones, LED displays, duplicated pedestrians, malformed
people, warped cars, distorted wheels, distorted architecture, excessive HDR,
plastic surfaces, fantasy lighting, oversaturation, collage, text overlay, and
watermark.

Ask for deep focus explicitly, and refuse bokeh by name. Left alone the model blurs the background into mush, which is the single clearest tell of a generated image. f/8, plus the whole list of blur words in the negatives, is what keeps the street looking photographed.

"Keep the immediate foreground empty." That empty pavement is where your characters go. Fill it here and there's nowhere to stage the scene.

Blank plate surfaces on realistic plate mounts. Removing the plates looks wrong; leaving them looks like a real car with unreadable text. Keep the mount, empty the surface.

Step 5 — The film

MODEL: GenX Pro -> Video → Seedance 2.5 Reference.

Settings: 9:16 · 720p · duration 15 seconds · audio on.

Attach your four sheets and address them by name in the prompt: the woman, the man, the dog, the background.

Create one 15-second, 9:16 photorealistic cinematic fashion-film sequence with
synchronized environmental audio.

@Woman is the complete approved woman sheet. Use its right panel for her exact
face and chestnut-brown hair; use its left/center panels for her burgundy trench
coat, cream blouse, dark skirt, shoes, proportions, and brown shoulder bag.
@Man is the complete approved man sheet. Use its right panel for his exact face
and dark wavy hair; use its left/center panels for his navy bomber jacket,
pale-blue shirt, charcoal trousers, shoes, and proportions. @Dog defines the
exact wire-haired terrier, white/tan/black markings, red-brown collar, and
proportions. @Background defines the deep-focus late-1980s avenue, warm
afternoon light, curbside black sedan, taxis, buildings, and pedestrian density.

Treat the sheets only as references. Never reproduce their triptych layout, gray
background, white dividers, or headless wardrobe displays. Recompose naturally
for vertical 9:16.

0-3s: The woman opens the sedan's curbside rear passenger door and steps
naturally onto the sidewalk, maintaining her outfit and handbag.
3-6s: Cut to the man approaching on the same sidewalk, naturally walking the
terrier on a relaxed thin brown leash. The dog walks correctly on four legs with
proper paw contact.
6-9s: Cut back as the woman walks away from the sedan. Her brown handbag slips
from her shoulder and falls naturally onto the pavement with realistic gravity.
9-12s: The man notices it; both bend and reach for the handbag simultaneously
with believable hands, fingers, posture, and body mechanics. The dog may remain
outside the tighter crop.
12-15s: Cut to a restrained close two-shot. They pause near the handbag and make
surprised, warm romantic eye contact without speaking. Hold the reaction for the
emotional payoff.

Fill the full 15 seconds through natural reaction holds, never artificial slow
motion. Preserve exact faces, hair, wardrobe, proportions, dog markings,
location, lighting, and screen direction across all cuts.

CAMERA LOOK: Premium late-1980s 35mm fashion commercial, emulating ARRI Alexa 35
dynamic range with Kodak Vision3 250D color. Use natural spherical-lens
perspective: approximately 50mm for street and medium shots and 85mm for the
final close-up, around f/5.6-f/8 so the environment remains recognizable. Stable
tripod or restrained dolly movement, realistic skin, warm highlight roll-off,
restrained grading, fine film grain, and natural motion blur only on movement.
No smartphone look or artificial bokeh. Vehicle plate mounts remain realistic
but completely blank.

AUDIO: Generate only synchronized natural ambience and effects: restrained
traffic, distant nonverbal pedestrian murmur, car door, footsteps, clothing and
handbag movement, quiet paws, leash/collar sounds, and a soft handbag impact.

AVOID voiceover, dialogue, speech, lip-sync, singing, generated music, prominent
barking, loud horns, readable text, brands, plate characters, captions,
watermark, identity drift, face/hair/outfit changes, dog changes or unexplained
disappearance, duplicated people, extra limbs/fingers/paws, malformed anatomy,
fused bodies, sliding feet, floating or changing handbag, warped cars/wheels,
camera jitter, flicker, frame warping, teleporting, modern objects, blurred
AI-looking backgrounds, plastic skin, excessive HDR, fantasy lighting, and
oversaturation.

Tell it what the panels are for, and tell it not to draw them. Two separate instructions. Without the second one the grey studio background and the white dividers leak into the film.

Timecode every beat. 0-3s, 3-6s. Prose like "then he notices" is unreliable across a fifteen-second take; a timecoded block isn't.

Ask for reaction holds, not slow motion. A model asked to stretch a short action will ramp the speed and it looks wrong. Ask it to let people pause instead.

No music from the video model. You want the ambience it generates — traffic, footsteps, the bag hitting the pavement — and you want to place the score yourself. Ask for one and you get both fighting.

Step 6 — The music

MODEL: GenX Pro -> Audio → Lyria 3 Clip Preview.

Trim the generated music to fit the video.

Create a 30-second original instrumental score for a vertical late-1980s
romantic cinematic fashion film about two strangers meeting unexpectedly on a
busy city sidewalk.

Style: elegant late-1980s romantic synth-pop film score, 82 BPM, B minor,
nostalgic, sophisticated, emotionally restrained, warm and hopeful. Use lush
analog synthesizer pads, a soft pulsing synth bass, delicate clean electric
guitar with subtle chorus, restrained electronic drums with light gated reverb,
gentle bell-like synth accents, and a simple memorable original melodic motif.
Premium cinematic stereo production with subtle tape warmth.

TIMELINE:
0:00-0:04 - Sparse atmospheric opening as a woman exits a black sedan. Warm pad,
gentle pulse, minimal percussion.
0:04-0:08 - Introduce a quiet rhythmic motif as a man and small dog approach
along the sidewalk.
0:08-0:12 - Pull back the arrangement and create delicate anticipation as her
handbag falls and both characters reach for it.
0:12-0:17 - Emotional melodic lift as they look at one another and hold
unexpected romantic eye contact. Warm, intimate, hopeful - not melodramatic.
0:17-0:20 - Resolve gently into a clean, elegant tail.
0:20-0:30 - Let the pad decay and fall away. This section runs past picture and
gets trimmed.

Instrumental only. No vocals, lyrics, spoken words, humming, chanting, dialogue,
or embedded sound effects. Leave room in the mix for existing street ambience,
footsteps, leash sounds, and the handbag impact.

Create a completely original composition. Do not imitate, quote, interpolate, or
reproduce any recognizable existing song, melody, recording, soundtrack,
composer, or artist. Avoid modern EDM drops, trap percussion, aggressive drums,
trailer music, comedy, horror, prominent saxophone solos, excessive
sentimentality, overly busy instrumentation, sudden volume changes, and abrupt
endings.

Trim to picture and lay it under the ambience, not over it. The traffic, the footsteps and the bag landing are what make the street real; the score is there to tell you what the look meant.

Give the music a timeline that matches the film's. A prompt that just says "romantic synth-pop, 20 seconds" gives you twenty seconds of wallpaper. Tell it where the lift goes and it arrives on the right frame.

Make it your own

Log in to GenX Pro or create a free account, then open the App and try these prompts.