Meta Feed video ad spec checklist before uploading an AI clip

Meta's Facebook Feed video ad spec, as the page read on 2026-10-07, is a 4:5 aspect ratio, a recommended resolution of 1440 x 1800 pixels, a length from 1 second to 241 minutes, a file size up to 4 GB, MP4, MOV or GIF files, and H.264 compression with square pixels, a fixed frame rate, progressive scan and stereo AAC audio at 128 kbps minimum. An AI-generated clip usually misses the ratio first, so the practical check is to render or conform to 1440 x 1800 before upload.

The spec as Meta published it

The page we fetched focused on Facebook Feed. It mentioned other placements such as Stories and Reels but did not include their technical numbers in the text we received, so this checklist is for Feed only.

Where AI clips usually miss

Most Sume video models do not list 4:5 as an aspect ratio. Seedance accepts 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16, and Gemini Omni Flash accepts 16:9 and 9:16, so a native 4:5 clip is not something to expect from /v1/videos. The standard fix is to generate at 3:4 or 9:16 and conform the result to 4:5.

Timeline 1.0 is the documented conform step. Its output width and height can be any even integers from 256 to 2160, so 1440 by 1800 is allowed, and the default fit is cover, which fills the frame and crops the overflow. Its output is an MP4. You choose fps from 24, 25, 30 or 60, or omit it to match the sources.

Render to the spec

The request below builds a 1440 by 1800 MP4 from a hosted clip. The audio spine is required, so for a clip that carries its own sound you detach the audio first or use a spine that matches. The source must already be a media.sume.com artifact, and Timeline needs an Idempotency-Key.

Call POST /v1/timeline-1.0/plan first. It is unbilled and returns the duration, segment count and billable minutes.

curl -sS -X POST https://api.sume.com/v1/timeline-1.0/plan \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "audio": {"mode": "silence", "duration_seconds": 6},
    "video": [{"source_url": "https://media.sume.com/artifacts/artf_demo/clip.mp4", "start": 0, "duration": 6, "fit": "cover"}],
    "output": {"width": 1440, "height": 1800, "fps": 30}
  }'

The pre-upload checklist

Run these checks on each variant before it goes into an ads manager. Use video inspect on a media.sume.com clip to read probe facts, or ffprobe on your own copy.

• Dimensions are 1440 x 1800, or at least above 120 x 120.
• The file is MP4 or MOV, not a stray format.
• Length is inside 1 second to 241 minutes, and file size is under 4 GB.
• Audio is stereo AAC at 128 kbps or higher; if the ad is silent on purpose, say so in your notes.
• Frame rate is fixed, not variable.
• Burned-in captions are legible inside the 4:5 frame, since cover crops the edges.

What we cannot promise

Sume documents the Timeline output as an MP4 but does not publish codec settings in the pages we used, so verify H.264 and AAC on a real output with a probe before you rely on it. Meta can change its spec, and the page reflects the date we read it.

Captions and sound

Meta's page says captions and sound are optional but recommended for video ads. For a feed placement where many people scroll with the sound off, burned-in captions are the safe choice. Sume's video captions route takes a public video URL and burns captions with a chosen style, at $0.20 for a clip up to 60 seconds. Do the caption step before the final conform if the captions are near the edge, or after it if you want them placed against the 1440 by 1800 frame.

Place the caption job after the Timeline render when you can. The caption design fields include placement anchors expressed as a fraction of frame height, so a 4:5 frame gets the line where you expect, not where a 9:16 frame would have put it.

If you produce many variants, apply the same conform step to all of them. A mixed set where half the clips are 9:16 and half are 4:5 cannot be compared fairly in a feed test, because the crop changes how much of the product is visible. Choose the placement first, decide the output size, and push every variant through the same Timeline body with only the source clip changing. That also lets you price the batch up front: with $0.10 per output minute, ten clips of 6 seconds is ten Timeline jobs and $1.00.