Energy drinks: the can still; no synthetic athlete VO
Pack shot plus captions — health-adjacent lines stay off the talking-head row.
Pack shot plus captions. Health-adjacent lines stay off the talking-head row.
An energy-drink ad dies when the can in motion is not the can on the shelf, or when a constructed athlete starts prescribing a protocol. The SKU is a still. Animate that. Put the flavour line in type you wrote. Do not buy a synthetic athlete voiceover.
The can still is the listing
If you would not print the still on a carton, do not buy motion. Front panel, colourway, the flavour you actually sell, the caffeine line legal already approved. Even light. Label readable. That frame is the contract.
AI product video is the product-hold job: photographs of the real object as ground truth, camera and scene allowed to move, the pack not allowed to re-typeset itself. Prompt motion only — slow orbit, condensation if the condensation is already in the still, a tab crack only if the tab is in frame. Pause on the flavour word. If it would fail a listing audit, throw the clip away. Do not "fix" it by generating a prettier can.
Lock the still before you spend on motion. Text-to-video will re-decide the tab, the flavour mark, and the milligrams every take. Labels are the expensive failure. A talking founder holding "your" can is a different row, and it is still not a reason to invent the SKU.
Do not put an outcome claim in the generate. "Improves focus," "safe pre-workout," "what you should drink before a PR" is a different legal file and a health-adjacent line. Describe the can. Leave outcomes off the picture.
Captions listen; claims are overlay
Words still ship. They ship as type, not as a mouth the model invented.
If a real named person — a founder, a brand athlete on payroll with a grant — actually spoke the line, add captions to a video transcribes that speech and burns styled subtitles. Captions listen. They do not invent an athlete. They do not translate. Set the language to the language spoken. Auto-captions fail on flavour names and milligram strings. Read the transcript. Fix the pack words by hand. Then burn.
If nobody spoke, do not run captions and hope a headline appears. Overlay prints the string you authored: flavour, a CTA, a legal line already on the can. Offers change; the can does not. Do not ask the product-video row to letter a claim into the shot. Generated glyphs misspell. Generated milligrams are not the milligrams on the carton.
Muted autoplay still needs the line on the file. A silent can orbit with no type is a pretty loop. A silent can orbit with the flavour you sell is the ad.
No synthetic athlete VO
Photoreal talking presenters on health are a YouTube Partner Program kill. Synthetic presenters on health and finance are a YPP kill: an AI persona styled as a human expert giving advice on health. Wellness remedies sit in YouTube's own example on purpose. A constructed sprinter walking "what you should drink for this workout," a generated dietitian with a can in hand, a talking "performance coach" — that is the third bucket. Labelling the face AI does not leave it. A labelled AI athlete is still an AI athlete.
A can is not a persona. A condensation bead is not a clinician. The kill is persona plus prescription.
Do not glue a talking protocol onto the pack shot inside one generate. Do not prompt native-audio rows for a VO that "sounds like a pro athlete." That is a voice you did not cast and a health line you will have to stand behind. If a founder or a real contracted athlete must speak, that person is on camera or on a mic you recorded, on a different day. Generation can still do the can. It cannot be the athlete.
Do not generate patients, residual illness, or a fake training lab. The still you photographed is the whole clinical set you are allowed: aluminium, liquid, label.
Gyms and supplement pages are nearby, not this can
Gyms sell member energy: peak-hour floor, a PR, the room. That page is occupancy on purpose. An energy-drink brand that posts a generated crowd "crushing a WOD" with your cousin can has used the gym playbook on a SKU that needs a label. Member energy is their footage. The can is yours.
Supplement brands convert on comparison: ingredient, dose, a label close-up of the powder. Useful neighbour. Not this listing. An energy drink is a can, a flavour word, a caffeine line. Copying the "this vs that protein" talking head onto a carbonated SKU is how you leave the pack and enter a health-advice persona. Overlay a flavour line on the locked can. Do not Omni a supplement explainer and call it the drink.
Do not rewrite those audience pages. This job is identity-hold the can, burn the line as captions or overlay, keep the synthetic athlete off the timeline. Pause on the flavour word. If it would fail a listing audit, the clip is not a SKU. Go back to the pack shot.
FAQ
Can we generate an athlete if we disclose that they are AI?
No. Disclosure does not move a video out of YouTube's third inauthentic-content bucket. A synthetic expert prescribing a pre-workout protocol is the pattern. Change the format: can still, type you wrote, or a real named person with a grant.
Should flavour and caffeine copy be captions or overlay?
Captions if someone actually spoke and you want that speech on screen. Overlay if you authored a line nobody said. Captions listen. Overlay prints. Neither job is a generate of an athlete.
Why not text-to-video a gym and composite our can later?
Because T2V re-decides the vessel, the liquid, and the room every take. The label you composite onto a cousin can is still a cousin can. Photograph the SKU. Hold it on the product-video door.
Does the gym page mean we should always show a workout?
It means occupancy and member energy beat equipment photography for that audience. For the drink, the can still is the SKU. A workout you filmed with a grant is allowed. A generated athlete VO is not.