Meta AI video ads from product photos: reported limits, Sume route

Meta is reported to turn up to 20 product photos into multi-scene video ads inside Advantage+, but that figure comes from a search-result snippet of a Leapbuzz article, not from a page I could open. Treat it as reported and confirm it in Ads Manager. Sume can build a similar clip, scene by scene, from your own photos.

The reported details, from the same snippet: 31 Advantage+ creative enhancements of which 17 are labeled AI, and 10 default-on enhancements including music on Reels ads. None of that is checked here.

What does Sume do with product photos?

Sume has two honest routes. The first is one clip per photo: POST /v1/videos with the photo in frame_images as a first_frame, so the clip starts from your real product shot. The second is a still-only video: Timeline 1.0 holds stills as static frames and joins them.

How do I join the scenes?

Generate one short clip per photo, then submit a Timeline 1.0 render with one video[] slot per clip and an audio.mode of silence or a music file. Slots can use transition types such as fade and dissolve. Output defaults to 1080 by 1920.

What does each step cost?

Clips bill at provider list times 1.25 by model, and Timeline costs $0.10 per output minute rounded up. Use GET /v1/videos/models for the live per-model price before you submit.

When might Meta's own tool suit you?

When you want Meta to choose scenes and publish in one step, and you accept its defaults. When you need to know which scene shows which product, or keep a record of every prompt, build it yourself.