Reels ads: 9:16, sound on, and a safe-zone brief for generated clips

Meta's Reels ads page says Reels ads use 9:16 and that Reels default to sound on. It reports that 9:16 video with audio and key messages kept in the safe zone got 2x delivery to Reels and 34.5 percent lower cost per result than image ads. That is Meta's own statistic, not an independent measure. For a generated clip, the brief is: vertical, audible, and nothing important near the edges.

What the page establishes

The page does not publish the safe-zone dimensions in the text this post relies on. Use Meta's preview overlay in Ads Manager to see the zones for your placement.

Turn the facts into a generation brief

Ask for aspect_ratio of 9:16 on POST /v1/videos, and where the model lists it, leave generate_audio on so the clip carries sound. The video docs say generate_audio defaults to the model's audio capability, and each model reports generate_audio in the catalog. A model that does not generate audio needs music or voice added from another route.

For a soundtrack, the Music Router generates music from a prompt, and the timeline surfaces can lay it under a clip. For a spoken line, the avatar video route turns a script into a talking video.

Prompt pattern

Describe the safe zone in the prompt in plain terms, since a model has no concept of Meta's overlay. The pattern below is a starting point to test, not a guarantee of placement.

{
  "model": "sume/auto",
  "prompt": "Vertical clip. Product centred in the middle half of the frame, lots of empty space at top and bottom. Natural ambient sound, a single line of upbeat music.",
  "aspect_ratio": "9:16",
  "duration": 8,
  "generate_audio": true
}

Checks before upload

Compare against an image ad on your own account instead of assuming Meta's figures transfer. The 34.5 percent number describes Meta's sample and may not describe yours.

• Play the clip with sound on and sound off; the opening second should make sense both ways.
• Overlay the Reels preview and confirm no key text sits under the interface.
• Keep the brand name inside the middle band.
• Add captions after generation for any spoken line, with the video captions route.

Sound design notes

Sound on by default changes the first second. A viewer hears the clip before reading it, so a distinct opening sound, a spoken first line or a short musical hit can do work that a silent clip needs text for.

Keep levels consistent between clips in a batch. A batch whose clips differ in loudness feels uneven when viewed in sequence.

Keep captions on every clip that carries speech, because many people still watch with the sound down in public. The video captions route burns captions onto an existing public video URL, with style, language and optional script alignment.