Meta's Reels safe zone checker: test a Sume frame with video-frames

Meta's Reels ads page points to a Reels safe zone checker offered in PowerPoint, Photoshop and Keynote formats. To test a finished video, export a handful of stills at the moments with the most text, then place them in the checker. Sume's video-frames endpoint returns up to 24 exact stills from one clip, at the source size.

What Meta's page says

The page "Instagram & Facebook Reels: Create Short Video Ads" (read 2026-10-05) recommends 9:16 vertical video and says Reels ads with key messages in the safe zone, quality audio and vertical video see better results. It lists a Reels safe zone checker for PowerPoint, Photoshop and Keynote to help confirm that key creative elements are in the safe zone.

The text I read does not give percentages or pixel margins, so this post does not either. Use Meta's checker as the source of truth for the margins.

Pull stills at the risky moments

Pick the times where text or a product label is on screen: the hook, any price card, the call to action. Video frames takes at[] (1 to 24 times in seconds) or an fps up to 2, a format of jpeg or png, and an optional max_edge from 16 to 2160. If you skip max_edge, frames keep the source size. The source must be a media.sume.com clip of 300 seconds or less.

curl -X POST https://api.sume.com/v1/video-frames \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: reels-safe-zone-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/reel.mp4",
    "at": [0.5, 3, 8, 14],
    "format": "png"
  }'

Read the result and use the checker

A submit returns 202 with a job id; poll GET /v1/video-frames/{id} until resource_status is ready. The frames array holds t, url, width and height for each still, as durable image artifacts. If one instant fails, that frame has a null url and the job does not fail.

Download the images and place each one in the Meta checker as the background layer. Move anything the overlay covers. Then fix it in Sume, not in the still: nudge the caption line with the caption design.placement.anchor_ratio, or re-run an overlay with a larger margin_ratio.

Cost and order of work

Video frames is billed by its own compute, not as a model: the docs say Modal container seconds times the Modal list price times 1.25, plus the platform fee, and never more than the hold reserved at submit. Frame extraction for a few stills is a small job, so run it on each candidate cut instead of only the final one.

Do this check before captions are burned in the last pass, so you can still move them.

• Use PNG for text-heavy frames.
• Keep to 24 frames per call.
• Do not trust one frame: check the first, the busiest and the last.