Perfume ad AI video: from one bottle photo to a short spot

Make a perfume ad with AI video: animate a real bottle photo with one move per shot, add a music bed, and check the glass and label in every frame.

5 min readSume
All posts

To make a perfume ad with AI video, start from a real, clean photo of the bottle as the first frame of each shot and ask for one slow move per shot: a push-in, a light sweep across the glass, mist drifting past, petals falling. Join three or four short shots over an original music bed, end on the bottle and the name, and check every frame, because glass, liquid and small label type are where generated video drifts from your product.

The Sume steps below come from the Video generation, Timeline 1.0, Music Router and Video frames docs, read on 2026-09-29. Limits marked as current behavior are read from Sume's code.

What are good perfume ad ideas?

A scent can't be filmed, so a perfume ad shows what the fragrance is made of or where it belongs. Ideas that keep the bottle on screen and the move simple:

  • Ingredients: citrus slices, rose petals or vanilla pods falling slowly behind the bottle.
  • Light: a warm highlight sweeping across the glass and the liquid inside.
  • Mist: a fine spray drifting through a backlit frame, then the bottle alone.
  • Place: the bottle on a stone ledge by the sea, on a dressing table at dusk, in snow.
  • End card: the bottle centered and still, with the name and where to buy it added in the edit.
  • Leave out famous faces and look-alikes, and claims about how long the scent lasts or how people react to it.

How do I make a perfume ad video from a bottle photo?

Animate the photo as the first frame of an image-to-video clip, one clip per idea: send its public HTTPS URL in frame_images with frame_type first_frame on POST /v1/videos. The request, the settings and the case for pinning a last frame too are the same as in AI video for jewellery and Keep a product exact in AI video. What changes for a fragrance is the prompt. The request has no camera setting, so name the one thing that moves and everything that stays fixed:

  • "Slow warm light sweep across the perfume bottle on black stone. The bottle, cap and label stay still and unchanged; only the light moves."
  • "A fine mist drifts through a backlit frame behind the bottle. The bottle does not move and the liquid level stays the same."
  • "Rose petals fall slowly behind the bottle on a dressing table at dusk. Static camera; the label faces the camera throughout."
  • generate_audio defaults to the model's audio capability. Send false when the music bed will carry the sound.

How do I add music and cut vertical and wide versions?

Generate an original bed with POST /v1/music-router/generate. Steer its length in the prompt ("a 20-second track"); a duration field is rejected. Then render the shots with Timeline 1.0 (POST /v1/timeline-1.0/render), which takes only this workspace's media.sume.com files, such as the clips and track Sume returned.

  • Use audio.mode "silence" to declare the length and add the music as soundtrack, or use the music file itself as the spine in audio.url.
  • In current code the render keeps only the spine and the soundtrack, so any sound a clip generated is dropped.
  • The default output is 1080×1920, vertical. For a wide cut, set output.width 1920 and output.height 1080. Video models offer ratios such as 16:9, 9:16, 1:1, 3:4 and 4:3, but no 4:5; for a 4:5 feed slot, render at 1080×1350 with the slot fit left at cover, the default, which fills the frame by cropping, so keep the bottle and label centered in the shot.
  • Put a fade or dissolve of up to 1 second between shots.

How do I check the bottle and label?

Only the frames you pin are your photo; everything between them is generated. Pull stills from each finished clip at chosen at[] times with POST /v1/video-frames, which is unbilled, and compare them with the product photo. For a fragrance, look at:

  • The label's text and logo, letter by letter: small type can warp.
  • The cap, the neck and the shape of the glass.
  • The liquid's color and level, which should stay the same through the shot.
  • Reflections that show a room or a hand that wasn't there.
  • Recut or regenerate any shot that fails.

How much does a perfume ad video cost?

You pay per step from your workspace USD balance: one video job per shot, one music track, and one render per cut.

From Video generation, Timeline 1.0, Video frames and the API pricing rate card, read 2026-09-29. Each paid rate is plus a 5.5% agent fee by default.
StepCallPrice
Animate each bottle shotPOST /v1/videosBy model, reserved at provider list × 1.25 per clip; see pricing_skus on GET /v1/videos/models
Music bedPOST /v1/music-router/generate$0.125 per audio
Join the shots (once per cut)POST /v1/timeline-1.0/render$0.10 per output minute
Stills to check the labelPOST /v1/video-framesUnbilled

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume