CPAP ads are the device still, not a sleeping generated face
SKU I2V. Do not generate a patient.
SKU I2V. Do not generate a patient. A CPAP ad that has to ship is the device you sell — hose, mask, console — moving just enough to prove it is a real object. A face on a pillow is a person. In this category that person reads as a patient, and a patient you invented is both a likeness problem and a testimonial you did not collect.
Photograph the unit. Animate the photograph. Leave people you do not have a grant for out of the bedroom.
Photograph the SKU, then buy motion
The mask is the product. The buyer already knows what their own face looks like. They need to recognise the interface, the headgear, the machine on the nightstand. Text-to-video will invent a prettier device with a melted logo. That is a failed SKU, not a cinematic upgrade.
Image-to-video keeps frame one as the file you sent. Seedance 2.5 is the still-to-motion family: four to thirty seconds, native audio, 720p on the I2V row, optional last frame when the shot must land on a second approved angle. Prompt only the camera. Do not prompt "a peaceful patient sleeping in 4K."
Use the manufacturer pack shot or a clinic photo of the actual unit. Do not generate a tablet, a hose colour, or an imprint the real SKU does not have. Wrong geometry on a medical device is a safety-adjacent miss, not a styling choice.
A sleeping face is a person, not a prop
"A 52-year-old man asleep with a CPAP mask" is not a product shot. It is a private figure. If you used a real headshot as a reference, it is a replica. If you invented a photoreal sleeper and put them in an ad, you still have a person who does not exist using a device, which is how a clip starts reading as proof.
A deepfake of a private person is not a style. TikTok's line on adult private figures used without permission is a ban, not a labelling suggestion. A generated patient in a sleep-clinic ad is that case wearing a hose. A standard photo release from a real patient does not automatically cover generating new footage of them; the grant has to say replica, specifically — a likeness release that covers AI generation.
A synthetic sleeper who "finally rests" is also a testimonial. The fake review rule does not care that the face was sampled rather than hired. Fabricated proof is fabricated proof. Real patients, with written consent, on camera, or no patients in the ad.
Minors are a hard stop. Do not generate a child on a machine. Do not "age up" a school photo. Do not put a generated kid in the background of a family bedroom.
What the ad is allowed to show
The device, from stills you own. Console, mask, tubing, travel case. I2V from those stills.
Empty room tone. Objects already in the plate. Not a recognisable private interior you do not have rights to.
On-screen specs as overlay. Model names and SKU strings belong in a text overlay you wrote. Video models do not typeset a medical SKU.
A real clinician, named, if they agreed to appear. Not a generated white coat.
Health claims stay on the approved labelling. The clip can show the object. It cannot promise that a generated sleeper's apnea resolved.
FAQ
Can I generate a generic sleeper if I never used a real photo?
A fully invented, non-lookalike face is not a private-figure replica. In a CPAP ad it is still a bad idea: it reads as a patient, it can read as a testimonial, and YouTube's expert-persona rules sit on health-adjacent talking heads. Film a real person with a grant, or show the device.
Is a mask-on close-up of a generated face safer because you cannot see the whole person?
No. A recognisable face is a likeness at any crop. A mask does not anonymise a replica. Use the product still.
Can staff appear if we already have a model release from the clinic shoot?
Read it. Use of photographs from a shoot is not permission to generate new sleeping performances. Get a replica clause or shoot the staff. Do not animate a receptionist's headshot into a patient.
Why not text-to-video a "lifestyle bedroom" and composite the SKU later?
Because T2V will invent a second machine in the room, or a face, or both. Start from the SKU still. If you need a room, shoot or generate an empty plate and overlay type. Do not ask a video model to invent the patient the still was meant to avoid.