Perfume: the flacon still; mist is Vidu if you want the hiss

Identity of the bottle first. Diegetic spray is a second row.

A perfume ad is two jobs pretending to be one clip. The flacon is identity: glass, cap, the wordmark that has to survive a pause. The mist is physics: atomiser, plume, the hiss that belongs to the spray you already photographed. Vidu Q3 Image to Video will write a mix. Asking it to invent the bottle and the hiss is how you get a cousin flacon with a great plume, or a perfect logo with a score that is not a spray.

The flacon is the SKU

A listing is a specific object. Shoulder, cap colour, the foil on this juice. Text-to-video “luxury perfume bottle, cinematic mist” invents a cousin bottle on a cousin marble. The brand that packed this flacon notices before the algorithm does.

Photograph the bottle. Same lens height you would use for a pack shot. Crop so the mark fills the frame if the ad is the SKU. Crop to the atomiser if the ad is the spray. Do not interpolate from a wet nozzle to a closed bottle of a different SKU and call the middle “our scent.”

Image-to-video starts from that frame. Frame one is the file. Labels, glass, and the cap survive because they were already in the pixels. The prompt is camera and physics — a slow orbit, a hold on the atomiser, a short plume. Re-describing “an amber bottle on black marble at golden hour” is how the model starts drawing a second bottle.

Skincare brands own a nearby CPG page: routine, face, a claim that is often health-adjacent. Jewelry brands own metal SKUs: stone, setting, no hiss. This post is not those pages rewritten. Perfume is the flacon. The job is the bottle you packed, then the spray if you need the sound. If the still’s bottle is already wrong, every second of mist is waste. Shoot again.

Mist is a second row

Vidu Q3 Image to Video turns a static image into a clip. Image required. Audio on. Durations are only 5s, 10s, and 15s. Max output 1080p. Sixteen credits listed; the row is marked discounted. No aspect-ratio list on this slug — crop to 9:16 or 16:9 before upload.

Native audio means soundtrack and frames from the same pass. Spray is the textbook case: the plume is visible, so the hiss has to be audible in the same denoise. Retrofitting a “spritz” sample onto invented mist is how the hiss lands on the backswing.

Write the mix into the 5s prompt. Name two or three sounds: atomiser click, hiss, glass settling. State “no speech, no music” if you want clean foley. Leave audio unspecified and Q3 invents a mix — including a murmur that does not belong on a pack shot.

A 5-second take is the shortest legal clip on this slug. If the plume fails in five seconds you do not have a cheaper 3-second retry here. Do not pad to 15s because the row is discounted. Audio is on. A “silent” brief still gets a band. This hiss is why you picked Vidu.

Nearby bottles are not this hiss

The AI product video generator is the wider product door when the flacon has to sit in a new set and the lockup must not. Reference-to-video holds the real shape while the scene moves. That is still identity. It is not the diegetic spray.

If the failure mode you cannot live with is a drifted wordmark, you are on an identity family, not on Vidu for resolution. 1080p is the Vidu ceiling. Prompt camera and mist only. The still already named the object.

Do not Veo a talking nose. Veo 3.1 is native audio, official 4K tier, 4 / 6 / 8 second clips. “Professional fragrance presenter, clean studio, top notes” is how you spend the talking meter on a persona. Spray is foley. The bottle is the product.

One generate, one job

Pick before you spend. The ad is the bottle: lock a still you would print, animate that file, camera only. The ad is the hiss: lock the atomiser still, run Vidu Q3 at 5s, write the spray into the prompt. You need both: two generates, then stitch. One model will not honestly own both constraints for a cheap test. If the wordmark grew a second serif, you used one still for two jobs.

FAQ

Can Vidu Q3 invent the bottle if I describe it well?

No. This slug requires an image. The still is the contract. A text-only “luxury perfume on marble” is a different family, and it will invent a cousin flacon. Photograph the bottle. Then decide whether this generate is for the mist.

Does a 5-second hiss need speech in the prompt?

No. Write the spray: atomiser, hiss, room tone, no speech. Native audio is on; a silent prompt still gets a soundtrack. Named foley is how you keep a murmur off the pack shot.

Why not one 15-second Vidu clip for bottle and mist?

Because 15 seconds is triple the shortest Q3 take, with a soundtrack attached, and discounted is not free. Prove the plume at 5s. Prove the logo on the locked still. Stitch if you need both.

Is this the same job as a skincare routine reel?

No. Skincare brands is face and claim. Perfume is a flacon SKU plus, if you want it, a diegetic hiss. Crop the bottle as a bottle. Do not put a generated clinician on the juice.