All resources
Prompt·5 min read·ChatGPT·Higgsfield·Seedance

The Coffee Spill Trend: Turn Two Photos Into a Luxury Sunglasses Ad (Higgsfield in ChatGPT)

The rain-soaked ad where the sunglasses fly off her face and the coffee hangs in the air. You take a photo of the person and a photo of the product, add Higgsfield inside ChatGPT, paste the prompt below, and Seedance 2.5 builds the whole 12-second shot as one take. The setup takes under 30 seconds.

What you need

Two photos and one plugin. That's the whole kit.

Photo 1, the person. Full body, in the exact outfit they'll wear in the ad, good light, face clearly visible. Take it without the sunglasses on. The model needs to know what the face looks like underneath, because there's a moment mid-shot where the glasses are off.

Photo 2, the product. The sunglasses on their own, on a plain surface, whole product in frame, nothing else in the shot. This is the pair the model locks to, so shoot the real thing, not a screenshot.

The video shows a third photo of a phone. You don't need it. The version below is just the glasses and the coffee, and it's cleaner.

Then the Higgsfield plugin inside ChatGPT. It puts Seedance 2.5, the video model this runs on, directly in your chat.

⚡ Add Higgsfield to ChatGPT

Use your own face or someone who has agreed to be the model.

Step 1: Add Higgsfield inside ChatGPT

Open the ChatGPT app, tap the plus button next to the message box, then Plugins. Search Higgsfield and add it. From then on, typing @ in the message box brings it up, and every generation comes back inside the chat.

  1. Tap here to get the Higgsfield plugin, then add it from the Plugins menu.
  2. Start a new chat so the two photos and the prompt land in one clean message.

Step 2: Upload the two photos

In the new chat, pick Higgsfield from the @ menu, then attach both photos in the same message. Person first, product second. The prompt calls them Image 1 and Image 2 in that order, and the order is what keeps the right face on the right body with the right glasses.

Step 3: Paste the prompt

Paste this under the two photos and hit enter. It tells Higgsfield to use Seedance 2.5, locks the person and the glasses to your photos, and builds the collision, the slow-motion inspection and the catch as one take. Boom. Twelve seconds, 9:16, one continuous shot.

Copy-paste this
GENERATION SETTINGS
Model: Seedance 2.5, through Higgsfield.
Duration: 12 seconds exactly.
Aspect ratio: 9:16 vertical.
Resolution: 1080p.
Output: ONE continuous, unbroken, live-action shot. No cuts.

STYLE
Ultra-photorealistic live-action luxury fashion film.
Real optics, real physics, real materials. Not CGI, not animation, not a product turntable.

REFERENCE IMAGES
Use exactly TWO image references.
Image 1 = the exact protagonist identity AND the exact outfit.
Image 2 = the exact sunglasses.
Everything else is generated from text.
No environment reference.
No coffee reference.
No wardrobe redesign.

# IMAGE 1: PROTAGONIST IDENTITY + EXACT OUTFIT
Use Image 1 as the exact identity and wardrobe reference.
Keep the same woman throughout:
same face,
same apparent age,
same skin tone,
same eyes,
same nose,
same lips,
same jawline,
same facial proportions,
same hairstyle,
same hair length,
same hair texture,
same body proportions,
same recognizable identity.
She wears the SAME OUTFIT shown in Image 1.
LOCK:
same garments,
same layers,
same colours,
same fit,
same materials,
same styling,
same shoes if visible,
same accessories belonging to the outfit.
Do not redesign her clothing.
Wide shot -> collision -> bullet time -> final shot = SAME WOMAN, SAME OUTFIT.

# IMAGE 2: EXACT SUNGLASSES
Use Image 2 as the exact physical sunglasses reference.
There is exactly ONE pair.
LOCK:
same frame silhouette,
same frame proportions,
same lens shape,
same lens tint,
same bridge,
same hinges,
same temples,
same thickness,
same materials,
same finish,
same hardware,
same visible details.
CRITICAL:
THE PROTAGONIST IS ALREADY WEARING THESE SUNGLASSES FROM THE FIRST FRAME.
They begin properly seated on her face:
bridge resting naturally on her nose,
both temples positioned correctly over her ears,
correct scale relative to her face.
Worn position -> collision -> release from face -> bullet-time suspension -> macro inspection -> catch -> placed back onto her face = THE SAME SINGLE PHYSICAL PAIR.
The sunglasses never appear on her lapel, in a pocket, or in her hand at the start.

# MASTER SCENARIO
Blue hour.
A bustling Manhattan-inspired luxury district immediately after rain.
The exact woman from Image 1 walks confidently down a broad New York sidewalk wearing the SAME outfit from Image 1 and the exact sunglasses from Image 2, already correctly worn on her face.
Her LEFT hand holds exactly ONE generic, unbranded takeaway coffee cup with a lightweight plastic lid.
Her RIGHT hand is empty and relaxed at her side.
A hurried pedestrian comes up the subway stairs.
The two make a small last-second avoidance move in the same direction.
The pedestrian accidentally clips the protagonist.
The same collision knocks the sunglasses physically from her face and sends the coffee cup out of her LEFT hand.
Time slows to approximately 2 to 3 percent of normal speed.
The CAMERA moves through the nearly frozen accident and inspects, in order:
1. the sunglasses
2. the coffee cup and its liquid
The objects never fly toward the camera. The camera physically travels to them.
Normal time returns.
The protagonist catches the exact same sunglasses and places them BACK onto her face.
The coffee falls to the wet sidewalk.
She immediately continues walking.
She does not retrieve the coffee.
She does not look back.
Tone: luxurious, confident, rebellious, mysterious, cinematic, controlled.

# MANHATTAN ENVIRONMENT
Generate an original Manhattan-inspired Midtown environment.
Include:
wide rain-darkened sidewalk,
wet asphalt,
yellow taxis,
dark town cars,
subway entrance,
black metal subway railings,
tall masonry facades,
glass storefronts,
Art Deco-inspired hotel entrance,
warm lobby lights,
cool blue-hour ambient light,
traffic signals,
restrained city reflections,
steam from a street grate,
natural background pedestrians,
cinematic urban depth.
No recognizable commercial branding in the environment.
No cyberpunk styling.
No impossible architecture.
Pedestrians stay on pedestrian surfaces.
Vehicles stay in traffic lanes.

# EXACT OBJECT COUNT
Exactly:
ONE protagonist matching Image 1.
ONE secondary pedestrian.
ONE pair of Image 2 sunglasses.
ONE generic coffee cup.
ONE lightweight plastic lid.
ONE coherent amount of dark coffee.
No phone. No bag. No duplicates.

# STARTING OBJECT OWNERSHIP
FROM THE FIRST FRAME:
FACE: wearing the exact Image 2 sunglasses.
LEFT HAND: holding the ONE coffee cup.
RIGHT HAND: empty.
The sunglasses are NOT on her lapel, NOT in a pocket, NOT in either hand.
Exactly one pair exists.

# 0.0 to 2.0s: CHARACTER REVEAL
Begin low over the wet sidewalk.
Track backward smoothly.
A taxi passes in the adjacent vehicle lane, creating restrained reflections.
The protagonist enters.
The camera rises gradually from:
footwear
-> exact Image 1 outfit
-> coffee in LEFT hand
-> face wearing the exact Image 2 sunglasses.
Clearly establish that the sunglasses are already being worn correctly.
Show enough of her face around the glasses to preserve the recognizable identity of Image 1.
Her hair, jawline, lips, cheeks and overall likeness stay clear.
She glances once at the street ahead and keeps walking.

# 2.0 to 3.7s: NEW YORK WALK
Continue the uninterrupted tracking shot.
Natural movement.
Same Image 1 outfit.
Same Image 2 sunglasses, correctly worn.
Coffee stays in her LEFT hand.
A secondary pedestrian comes up the subway stairs:
adult,
neutral stone-grey coat,
dark trousers,
empty-handed,
no sunglasses,
no coffee.
They glance briefly toward traffic.
Both people notice each other too late.
Both make a subtle avoidance move in the same direction.
Their paths converge.

# 3.7 to 4.4s: COLLISION
The pedestrian's ELBOW accidentally clips the protagonist's upper arm and upper torso.
At almost the same instant, the pedestrian's shoulder or coat sleeve makes a small, physically believable contact with ONE OUTER TEMPLE of the sunglasses she is wearing.
The collision is brief, accidental, restrained, non-aggressive.
The protagonist recoils backward about half a step.
Her shoulders rotate.
Her head reacts with a small inertial motion.
Because the sunglasses are physically contacted while her head moves:
the frame shifts sideways,
one temple slips free from behind one ear,
the bridge lifts away from the nose,
the opposite temple follows,
and the entire intact pair separates from her face.
This is one physically continuous removal event.
No teleportation. No instant disappearance. No second pair.

# SUNGLASSES RELEASE FROM FACE
The exact Image 2 sunglasses receive ONE short outward-and-slightly-upward impulse.
Release order:
outer temple receives contact
-> frame shifts
-> one temple clears one ear
-> bridge separates from nose
-> second temple clears
-> the entire pair becomes airborne.
They travel only 20 to 35 cm from her face.
They begin one restrained rotation caused by the off-centre impact.
They do NOT launch toward the camera.
The instant the sunglasses fully separate: FACE = NO SUNGLASSES.
Her uncovered face stays unmistakably Image 1.

# COFFEE RELEASE
The SAME coffee cup leaves her LEFT hand.
It receives one short outward and downward rotational impulse.
The cup tilts and rotates.
Because of the rotation and the liquid's inertia:
ONE edge of the lightweight lid partially lifts.
The opposite side stays attached.
A REAL gap forms.
ONLY AFTER the gap exists does coffee begin escaping.
Initial coffee: one restrained dark ribbon, a small meniscus, several connected droplets.
Immediately after release: LEFT HAND = EMPTY.

# FACE AND HAIR CONTINUITY
Once the glasses leave her face, the protagonist is still unmistakably Image 1.
This is critical.
Do not generate a different face after removing the glasses.
Maintain the same eyes, eye spacing, eyebrows, nose, lips, jawline, skin and hairstyle.
Her collision expression is restrained surprise.
No screaming. No exaggerated face. No beauty transformation. No outfit change.

# 4.4 to 5.2s: ENTER BULLET TIME
After the sunglasses have fully left her face and the coffee has left her LEFT hand, smoothly slow the entire physical world to approximately 2 to 3 percent of normal speed.
Not absolute zero motion.
Residual movement remains in the sunglasses rotation, her hair, her clothing, the coffee cup, the coffee liquid, the pedestrians, the traffic and the steam.
The camera leaves its original path and begins a controlled three-dimensional arc.
Establish the complete frozen event.
Spatial arrangement:
SUNGLASSES: near her face, shoulder height, slightly to one side.
COFFEE: lower and to her left.
FACE: uncovered.
LEFT HAND: EMPTY.
RIGHT HAND: EMPTY.

# 5.2 to 7.4s: SUNGLASSES INSPECTION
The CAMERA physically moves toward the exact Image 2 sunglasses.
THE SUNGLASSES DO NOT APPROACH THE CAMERA.
First show the entire suspended pair: both lenses, full frame, bridge, both hinges, both temples.
Then pan in closer.
Extreme photographic macro.
Preserve the exact Image 2 geometry.
Show real frame material, lens tint, lens thickness, hinge hardware, tiny scratches, dust, subtle fingerprints, realistic city reflections.
A lens may briefly reflect the protagonist's uncovered face. That reflection matches Image 1.
Do not redesign the glasses.
Do not rotate them artificially toward the camera.
When the camera moves away, the sunglasses stay where they are in world space.

# 7.4 to 9.4s: COFFEE INSPECTION
The camera physically travels to the SAME coffee cup.
The sunglasses stay behind.
THE CUP DOES NOT APPROACH THE CAMERA.
First show the complete cup: its tilt, its rotation, the attached lid, the one lifted lid edge, the real gap, the escaping coffee.
Then pan in to macro.
Show paper fibres, plastic texture, the wet rim, lid deformation, the coffee meniscus, surface tension, small bubbles, connected droplets.
Coffee exits ONLY through the physical opening.
No giant splash. No coffee passing through a solid lid.
Cup, lid and liquid stay anchored when the camera leaves.

# 9.4 to 10.2s: RECONSTRUCT THE TABLEAU
The camera physically pulls back out.
Show the complete frozen accident at once:
the Image 1 protagonist in the same Image 1 outfit,
uncovered face,
empty LEFT hand,
empty RIGHT hand,
the Image 2 sunglasses,
the coffee cup,
the lifted lid,
the escaping coffee.
All objects stay on their continuous trajectories.
No reset. No teleportation. No object followed the camera.
Prepare the composition for a believable sunglasses catch.

# 10.2 to 12.0s: TIME RETURNS
Return smoothly to normal speed.
NO CUT. NO RESET. NO NEW IMPULSE.
Every object continues with the exact position, velocity, orientation, direction, rotation and momentum it already had.

# SUNGLASSES CATCH AND RE-WEAR
The exact Image 2 sunglasses continue along their existing trajectory.
Her RIGHT hand, empty throughout, moves naturally into their path.
Show continuous physical contact:
empty fingers approach
-> fingertips touch ONE temple
-> fingers close
-> the same sunglasses are now controlled by her hand.
No magical catch. No teleportation.
She immediately returns the exact SAME sunglasses to her face:
held temple moves toward her face
-> bridge aligns with her nose
-> one temple clears hair and ear
-> second temple clears the opposite ear
-> bridge settles naturally
-> the sunglasses return to the same worn position established at the start.
Her Image 1 facial geometry is unchanged.
No clipping through hair or ears. No resizing. No sunglasses transformation.
The final worn pair looks identical to the pair worn in the opening shot.

# COFFEE IMPACT
The SAME coffee cup lands on the wet pavement.
The existing lid flexes further on impact.
Coffee spills naturally.
No exploding cup. No infinite liquid.

# FINAL SHOT
The protagonist immediately regains composure.
She is wearing the SAME exact Image 1 outfit AND the SAME exact Image 2 sunglasses from the opening shot.
She continues walking deeper into Manhattan.
She does not retrieve the coffee.
She does not look back.
Final composition:
FOREGROUND: the coffee cup on wet stone with its restrained spill.
BACKGROUND: the Image 1 woman walking confidently away, same outfit, same sunglasses properly worn.
Yellow taxis move naturally in their lanes.
Steam catches the city light.
Pedestrians keep moving.
Finish with a restrained optical rack focus: COFFEE CUP -> PROTAGONIST.
End while she is still walking.

# NON-NEGOTIABLE CAMERA LAW
The bullet-time illusion is created by EXTREME WORLD SLOWDOWN + CONTINUED CAMERA MOVEMENT.
Objects never fly toward the camera for inspection.
The camera physically moves:
wide tableau
-> sunglasses
-> coffee
-> wide reconstruction
-> the catch.
Each move shows the entire object first, then pans in to the macro, then pulls back out.
No camera teleportation.
No hidden cuts.
No camera passing through bodies, props, liquid, architecture or pavement.

# PHYSICAL CAUSALITY
Every event has a physical cause.
The elbow causes the body recoil.
Sleeve or shoulder contact causes the sunglasses to slide physically from her face.
Grip release causes the coffee to leave her LEFT hand.
Cup rotation causes liquid inertia.
Liquid inertia lifts one lid edge.
The real gap lets the coffee escape.
Off-centre contact causes the sunglasses to rotate.
Camera movement creates the macro scale.
Her hand intercepts the sunglasses.
Gravity brings the cup down.
Nothing transforms just because the camera distance changes.

# PRIORITY ORDER
1. exact Image 1 facial identity
2. exact Image 1 outfit
3. Image 2 sunglasses worn correctly at the start
4. Image 2 sunglasses geometry
5. physically believable sunglasses release from her face
6. uncovered Image 1 face after release
7. anatomically correct hands
8. empty hands after object release
9. continuous trajectories
10. correct coffee physics
11. realistic Manhattan environment
Sacrifice background complexity before sacrificing identity, outfit or sunglasses continuity.

# STRICT NEGATIVE CONSTRAINTS
NO sunglasses on her lapel, in a pocket or in her hand at the start.
NO protagonist without sunglasses at the start.
NO second pair of sunglasses.
NO sunglasses instantly disappearing or teleporting off her face.
NO sunglasses clipping through her head.
NO sunglasses breaking or changing geometry.
NO sunglasses magically returning to her face.
NO outfit change. NO wardrobe redesign.
NO face change after the sunglasses leave. NO identity drift.
NO phone. NO bag. NO extra props.
NO coffee reference image. NO branded coffee cup.
NO exploding lid. NO coffee passing through plastic.
NO object flying toward the camera.
NO hidden cuts. NO jump cuts.
NO CGI product turntable. NO superhero physics.
NO cyberpunk New York.
NO text, captions, logos or title overlay.

Generate the entire sequence as ONE continuous, physically coherent, ultra-photorealistic 12-second Manhattan fashion-film accident using Seedance 2.5, 9:16 vertical, 1080p.

Swap the product

The prompt is built around sunglasses because that's what the viral version used, and the catch at the end only works with something she can wear again. For a drink, a bag, a watch or anything else, change three things and leave the rest alone.

  • In the Image 2 block, describe your product instead of the glasses. Keep the same lock list, same silhouette, same materials, same finish.
  • In the release beats, the product leaves her hand instead of her face. One impulse, one rotation, stays close to where it was released.
  • In the catch, she catches the product and holds it up instead of putting the glasses back on.

Keep the coffee, the collision and the camera law exactly as written. Those three are what make it read as real footage instead of a render.

The things that break it

Sunglasses photographed on someone's face. The model reads the face as part of the product and you get two people in the shot. Shoot the glasses alone on a table.

A product photo with half of it cropped out. Whatever is missing gets invented, and it will not match the pair you're selling.

Trimming the negative constraints to make the prompt shorter. They look like padding. They're the reason you get one pair of glasses instead of two, and one continuous shot instead of three.

Asking for a cut. The moment you write "then cut to" anywhere, the slow motion falls apart. The whole effect is the camera moving while the world stops.

Get the next one first

New prompts every week.

Free. The new drops and the tools behind them, before they hit the feed.

No spam · New issues Sunday · Unsubscribe anytime

Need it custom?

Want this built for you?

Tell me the idea and I’ll build it. An app, a tool, an automation. You don’t need to be technical.

Frequently asked questions

Adding it to ChatGPT is free. Generations use Higgsfield credits, the same ones as the site, and a 12-second 1080p Seedance 2.5 render costs more than a short clip. So get the plugin here, read the prompt once, and run it when your two photos are right.

Seedance 2.5, through the Higgsfield plugin. The settings block at the top of the prompt tells it so, you don't pick anything by hand. It's the same model behind the door to anywhere effect.

No. Whoever is in Image 1 becomes the protagonist, so it can be you, a friend, or a model you've hired. Use someone who has agreed to be in the ad.

Two pairs usually means the person in Image 1 was already wearing glasses, so the model counted them twice. A face that drifts usually means Image 1 was small, filtered or side-on. Reshoot Image 1 without glasses, front on, in daylight, and keep every line of the identity lock in the prompt.

Yes, see the swap section above. Change the product block, the release beat and the catch, keep everything else. If you want a still campaign to go with it, the ChatGPT photoshoot runs on the same plugin.