Can AI make an App Store preview video? Apple's rules

No: a video generator can't make the App Store preview itself. As of 2026-09-29, Apple's page says app previews use "footage captured on device" and must "show only content within the app itself". A generated clip invents its own pixels, so use it for ads and social posts, not for the preview slot.

Apple's numbers below are from its own app preview specifications and app previews guidance, read 2026-09-29. The Sume facts come from Video generation and Timeline 1.0.

What are Apple's app preview specifications?

Apple lists one set of file rules for every preview. Resolutions differ per device family, so read the device table on Apple's page before you export.

Why can't a generated video be the preview?

Apple's guidance is about what the footage is. Previews "demonstrate the features, functionality, and user interface of your app using footage captured on device", and Apple asks that transitions "don't imply functionality that your app doesn't include". A video model draws a plausible screen, not your build. It can misplace a button or invent a feature.

Apple also says to capture footage from the device with QuickTime Player on macOS. Sume has no step for that: its media tools, Timeline included, read only files already hosted on Sume's media.sume.com, so a screen recording of your app is not something to route through it.

What can AI make for an app instead?

Everything around the preview: a short social or ad clip, an intro before the real screen capture in your editor, or a music bed. Apple's copy notes that previews play muted by default and suggests text for context, and that a single soundtrack keeps scenes continuous. Those tips shape the clip you assemble yourself.

For a promo clip, start from a screenshot or a key visual as the first frame and let the model animate a scene around it. Then treat the result as illustration: check every frame against the real app before you publish it anywhere, and don't let it stand in for the UI.

• frame_images with frame_type: "first_frame" makes an image-to-video request; the image URL must be public HTTPS.
• aspect_ratio: "9:16" gives a vertical clip on models that list it in supported_aspect_ratios.
• GET /v1/videos/models shows each model's durations, resolutions and aspect ratios before you spend.

curl -X POST "https://api.sume.com/v1/videos" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2",
    "prompt": "A phone on a desk, soft daylight, slow push-in on the screen",
    "aspect_ratio": "9:16",
    "duration": 5,
    "frame_images": [{
      "type": "image_url",
      "image_url": { "url": "https://example.com/hero-screen.png" },
      "frame_type": "first_frame"
    }]
  }'

How do I join clips into one vertical file?

Timeline 1.0 joins Sume-hosted clips into one MP4. The default output is 1080×1920, transitions on later slots can be fade or dissolve (the kind Apple calls straightforward), and renders are billed per output minute (see API pricing). A Timeline render is not an app preview: its length, resolution and content still have to meet Apple's table, and the footage still has to be captured on device.

For a logo intro or end card, see the existing post on a logo animation API.