Facebook Video Ad Structure: Specs, Length & the 3-Act Hook (2026)

There is a stubborn belief that a good Facebook video ad is just a shorter version of a TV spot: open on the brand, build a little story, land the logo at the end. That instinct made sense when people watched ads with the sound on and the patience to wait for a payoff. It does not survive contact with the modern feed. Facebook video ads now compete inside a ranking system that scores attention in fractions of a second, plays muted by default, and decides how much it likes your creative long before a human ever forms an opinion. Understanding how that system reads a video is the whole game, and it changes both how you cut the ad and how you spec it.

How Meta's ranking actually reads a Facebook video ad

Meta does not rank your ad on whether it is good. It ranks on predicted engagement and predicted conversion, then runs an auction where your effective cost is shaped by how confident the system is that people will watch and act. Video gives that predictor far more to work with than a static image. A still frame produces a single yes-or-no scroll decision. A video produces a curve: a thumb-stop, a hold, a drop-off, a completion. The model samples that curve early and uses it to decide how aggressively to spend your budget.

This is why early retention dominates everything else. If a meaningful share of viewers bail in the opening moment, the predicted-engagement signal sags and delivery throttles. A six-second clip watched to the end can out-deliver a thirty-second clip abandoned at second two, because Meta cares about the ratio of watched to served, not the absolute runtime. The practical consequence is that the opening of a Facebook video ad is doing two jobs at once: persuading a person and feeding a classifier. Both want the same thing, which is for something legible and interesting to appear immediately.

The 2026 reality: Reels, Advantage+, and muted autoplay

Three concrete shifts make 2026 different from the era that produced the old "brand-first" habit, and each one has a direct effect on how you build creative.

The first is placement gravity. Reels and short-form vertical video now make up the majority of time people spend inside Meta's apps, and a large slice of ad inventory has followed that attention. Building a campaign that only ships a 4:5 feed cut leaves Reels and Stories inventory to be filled by an awkward auto-crop of the wrong asset. The second is Advantage+ and Advantage+ Shopping Campaigns (ASC), where Meta leans on automated placement and creative selection rather than hand-tuned ad sets. When the platform is choosing placements for you, your single biggest lever stops being the bid and becomes the spread of distinct creatives you feed it. The third is the oldest and most ignored: autoplay is muted. The sound-on experience is the exception, so a Facebook video ad that depends on a voiceover to make its point is, for most impressions, a silent film with no subtitles.

These are not stylistic preferences. They are the operating conditions every Facebook video ad runs under, and they dictate the specs and the structure that follow.

Facebook video ad specs by placement

One of the most common mistakes is treating "Facebook video ads" as a single spec. They are not. Feed, Reels, Stories, and in-stream are distinct placements with distinct shapes, and what looks clean in one gets cropped or letterboxed in another. Here is how to think about each.

• Feed: a 4:5 vertical fills the most screen height without triggering a crop, while 1:1 square is the safe universal fallback. Keep critical text and your CTA inside the central area; the very bottom of a feed video sits near the caption and the "Sponsored" label, so a button parked in the last few percent of frame height gets visually crowded.
• Reels: full-screen 9:16. This is where safe zones bite hardest. Reserve the top roughly 14 percent and the bottom roughly 35 percent of the frame for the interface, because the username, audio attribution, caption, like and share icons, and the CTA chip all overlay that lower band. Anything important in those zones is partially hidden behind buttons.
• Stories: also 9:16 and full-screen, but the safe band is different. Leave the top and bottom roughly 14 percent clear; the profile and progress bar live up top and the swipe or CTA affordance lives at the very bottom.
• In-stream: 16:9 horizontal, served before or during longer video. This is the one placement where the old landscape, sound-likelier-on assumptions partly hold, but it is a small slice of delivery and should never be your primary cut.

The point is not to memorize percentages. It is to design with the seams in mind so your hook caption, your demonstration, and your CTA all sit inside the visible core of every placement you run. A reframe per placement beats one asset stretched across all of them; our walkthrough on producing 9:16 video that fills the frame covers the vertical case in depth.

Facebook video ad length in 2026

Length is the question people most want a number for, so here is a usable range rather than a hedge. For cold prospecting, most Facebook video ads that perform land between 15 and 30 seconds, with the strongest hooks often resolving the core promise inside the first 6 to 10 seconds in case that is all anyone watches. Reels-native cuts tend to run a touch shorter and snappier than feed cuts. In-stream tolerates more length because the viewer has already committed to watching something.

The deeper truth is that retention ratio governs outcomes more than runtime. Extending a video only helps if the curve holds; if your thirty-second cut bleeds viewers after second eight, a tight twelve-second version of the same idea will usually deliver cheaper. Start short, watch where the curve falls off in the breakdown, and only add length where people are clearly staying.

The three-act structure that beats static on cold audiences

Once you accept that the system reads a curve and most viewers are watching muted on a vertical screen, a structure falls out almost mechanically. The Facebook video ads that hold up on prospecting traffic tend to move through three acts, each doing exactly one job, in the order the auction samples them.

Act one: the hook, first 0 to 3 seconds

The job here is to stop the scroll and frame the situation, with no brand intro and no slow establishing shot. Open on the viewer's problem or the desired outcome, shown plainly. Put a legible caption on the very first frame, because in a muted feed the on-screen text is what communicates; treat this as a default, the way our guide on captioning video ads for sound-off feeds lays out. If you are short on opening lines, the patterns in our collection of scroll-stopping hook formulas are a good place to start.

Act two: the demonstration, roughly 3 to 20 seconds

The job is to make the claim concrete and believable by showing the thing happen, not by listing features. Use two or three quick proof beats rather than one long explanation, because each cut resets attention and keeps the retention curve from sagging. Match the visuals to the caption so the screen always shows what the text claims.

Act three: the offer and CTA, the closing seconds

Only now do brand, price, and button belong on screen. State the action as a verb, repeat the core promise once in compressed form for anyone who joined late, and hold the end card long enough to actually read. A button that flashes for half a second converts nobody. For wording, our list of CTA formulas worth keeping on hand covers this beat.

This ordering works because it mirrors how delivery samples the ad: the hook earns the watch, the demonstration earns the trust, and the offer earns the click. Lead with the offer instead and you spend your most expensive frames on the part a cold viewer cares about least. If you want the same logic at length, our framework for writing ad scripts walks through these beats line by line.

A worked example for a DTC skincare ad

Picture a vitamin-C serum brand running cold prospecting. The hook opens on a close-up of dull, uneven skin with a caption reading "Three weeks. Same person. No filter." The demonstration cuts between the morning routine, a finger swiping the texture, and a genuine before-and-after under matched lighting, each shot held just long enough to register. The offer arrives last: the product name, the price, a single line of proof such as a return rate or a reorder figure, and a verb-led button like "Start the 3-week test." The brand logo shows up for the first time about twenty seconds in, by which point the viewer already has a reason to read it. Notice there is no fictional gadget and no bracketed template to fill in. The structure is fixed; the specifics come from your actual product.

Why one Facebook video ad is never the whole strategy

Structure gets a single ad to perform. Sustained performance comes from how that ad multiplies, and 2026's delivery mechanics reward this specifically. Under Advantage+ and dynamic creative, Meta pools your assets and tests combinations, then consolidates spend onto the winners rather than splitting it evenly. Feed it one creative and the system has nothing to choose between, so it cannot do the job it is best at. Feed it a deliberate spread and it finds the pocket your audience responds to faster than any manual rotation would.

The discipline that makes pooling pay off is varying one element at a time so each test teaches you something:

• Hook variants: the same body with three different opening three-second beats. This is the highest-leverage variable, and the question of what to vary first when testing video creative is its own discipline.
• Placement reframes: a 9:16 cut for Reels and Stories, a 4:5 or 1:1 cut for feed, a 16:9 cut where in-stream applies. The same idea recomposed for the placement, not stretched to fit it. Choosing between formats up front is its own decision, which our piece on carousel versus video on Meta unpacks.
• Angle variants: price-led, outcome-led, social-proof-led. Same product, different reason to care.

The constraint that stops most advertisers here is not strategy but production. Cutting ten genuinely distinct Facebook video ads by hand is days of work, so small advertisers ship one and hope. The economics only function when a variant is cheap enough that retiring it after twenty dollars of spend costs nothing, which is why how fast you can iterate on video tends to decide who wins a Meta account.

A pre-launch check for Facebook video ads

Before a video goes live, walk it against the failures budget cannot fix:

• The caption is legible on frame one at phone size, held at arm's length.
• The ad makes sense with the sound off. Watch it muted before you trust it.
• The hook names a problem or outcome in the first three seconds, not your brand.
• The aspect ratio matches each placement: no letterboxing in Reels, no CTA cropped out of feed.
• Text and CTA sit clear of the placement safe zones, especially the lower band in Reels and Stories.
• The CTA is a verb and the end card holds long enough to read.
• At least three hook variants are queued, not one.
• The landing page promise matches the ad's promise closely enough that the click is not a bait-and-switch.

Frequently asked questions

What are the Facebook video ad specs in 2026?

Plan around placement rather than a single export. Use 4:5 or 1:1 for feed, 9:16 for Reels and Stories, and 16:9 for in-stream. Keep important content inside the central safe area, and for Reels in particular reserve the lower third of the frame for the interface overlay so your CTA and captions are not hidden behind buttons.

Do Reels ads need a different cut than feed ads?

Yes. A Reels ad is full-screen 9:16 with a heavy interface overlay on the bottom, so a feed cut dropped into Reels either gets letterboxed or has its lower content swallowed by the UI. Recompose the same idea vertically and shift key text and the CTA up into the visible band rather than stretching the feed asset.

Do Facebook video ads really clear the auction cheaper than static?

On cold prospecting audiences, video frequently wins more impression share than a matched static at a comparable or lower effective CPM, because the system expects it to hold attention. On warm retargeting audiences the picture flips: a clean static showing the offer and price often converts as well or better, since those viewers need reminding, not convincing.

How many Facebook video ad variants should I launch?

Enough to let delivery sort, few enough to learn from. Three to five genuinely different creatives per ad set is a sensible floor, varied by hook, placement, and angle rather than the same clip recut. Retire the weak ones quickly and let spend consolidate onto whatever the system favors.

If the wall between you and a real testing spread is production time, that is the narrow problem Aitachyon is meant to solve: it turns a single product into multiple captioned video cuts, framed for feed and Reels, so you can launch a structured spread on day one instead of shipping one ad and hoping the auction is kind.