Facebook Video Ad Examples: 7 Campaigns Meta Documented, and How to Rebuild Each One

On mobile Feed, your headline may never appear

A team ships a Facebook video ad with a sharp headline, a tight description and a Shop Now button, then buys awareness with it. On Facebook Feed on mobile, none of those three things renders. Meta says so on the spec page for that exact placement: select awareness as the video ad objective, or a performance goal like maximising reach or two second continuous video views, and the ad footer on the video ad, including the headline, description and call-to-action button, will not be displayed on Facebook Feed on mobile devices. The video is the entire ad.

That gap is why most Facebook video ad examples are useless to study. A screenshot with no date, no objective and no placement teaches you a vibe. So this guide only uses campaigns Meta published itself, which means every example has an address you can open. Each gets the same treatment: what the opening seconds do, the format envelope it was built inside, and one step to rebuild the mechanic for your product.

The seven, in one line each:

• Instamart. The identical creator video, run as a partnership ad instead of a standard video ad.
• bonprix. Twenty two creator cuts cast across body types, instead of one hero film.
• Boden. Vertical, sound on, message inside the safe zone, shipped as a four day burst.
• BYD. Text overlays that make the ad legible in the first seconds.
• Rexona Australia. Generated landmarks and local slang that only one country reads correctly.
• HUGO BOSS Middle East. Creative iterated by generation, from product images and brand guidelines.
• GUMAC. A Feed video whose next step is a Messenger thread, not a landing page.

If you want the layer by layer reading method rather than the campaign list, our breakdown of Instagram ad examples sets out the four questions to ask of any ad. This piece is the Facebook counterpart, and it is more concerned with the box each ad had to fit inside.

The format envelope every Facebook video ad is built inside

In Meta's own words, the Ads Guide provides design specifications and technical requirements across each format and placement. That last phrase does the work. There is no single set of Facebook video specs, there is a grid of them, and the cell you land in depends on the placement and the objective you picked.

Design recommendations are not technical requirements

The Feed video page splits into two labelled blocks, and they are not equally binding.

• Design recommendations: MP4, MOV or GIF, a 4:5 ratio, 1440 x 1800 pixel resolution, with video captions and video sound both listed as optional but recommended.
• Technical requirements: 1 second to 241 minutes, 4 GB maximum file size, 120 pixels minimum width and height.

Only the second block rejects an upload. A 1:1 video is not blocked from Feed, it is off recommendation. And a window running from one second to just over four hours constrains nothing, which is why "how long should a Facebook video ad be" has no answer on the spec page. Recommendations describe where the placement looks native. Requirements are the floor.

The genuinely tight number on that page is a text one. Meta lists primary text at 50 to 150 characters and the headline at 27 characters. Twenty seven characters is roughly four words, and on an awareness buy on mobile they may not render at all.

The same page said something different in 2024

It is easy to read today's numbers as permanent. They are not. A May 2024 capture of the same guide recommended 1:1 for desktop or mobile, or 4:5 for mobile only, at a resolution of at least 1080 x 1080 pixels. Today it recommends a flat 4:5 at 1440 x 1800. The technical requirements never moved: the 2024 capture already read 1 second to 241 minutes.

Two consequences follow, and they matter more than the numbers:

• Build to what the page says the week you build. Not to a spec sheet you saved last year.
• Distrust any ratio quoted without a date. That includes the older ad you are admiring, which was made under different guidance than you would get today.

What the placement actually changes

Two placements in the same guide, two different masters if you want either to look native:

Placement	Recommended ratio	Recommended resolution	Duration limit
Facebook Feed	4:5	1440 x 1800	1 second to 241 minutes
Instagram Feed	9:16	1080 x 1920	1 second to 60 minutes

Same ad account, same footage, two crops. If you have wondered why a clip that looked right in one placement looks stranded in another, this table is most of the answer, and the same logic drives the differences we cover in TikTok versus Instagram ads.

1. Instamart: the same video, sent from a different handle

Instamart is an Indian quick commerce platform. For the sales week around Valentine's Day it sourced local nano creators who specialise in authentic lifestyle content, and asked them to develop contextually relevant storytelling concepts about fast gift delivery.

What the test actually isolated

This is the rare case study that isolates one variable properly. Meta states that the creative, the audience and the budget were identical across both cells, so the only difference was the ad format. One cell ran the nano creator clips as standard video ads. The other ran the same clips as partnership ads, which deliver from each creator's handle as well as the brand's. Meta reports 25.4% more in app subscriptions and a 17% lower cost per in app subscription.

Nothing about the film changed. The name above it did.

Rebuild it: treat distribution as a creative variable

Most teams test hooks, thumbnails and lengths, and treat the sending handle as fixed. It is not. Before you commission a fourth variant of a clip that is already working:

• Take the clip that is already performing and change nothing about it.
• Run it through a second distribution path, from a creator account rather than the brand account.
• Hold creative, audience and budget constant, so the format is the only difference.

It costs one campaign setup and no new production. The lesson generalises past partnership ads: the question "who appears to be saying this" is a creative decision, which is the same reason UGC style ads outperform brand film for so many categories, and why the creator-led ad examples worth studying are rarely the most polished ones.

2. bonprix: 22 cuts, cast for the buyer

bonprix is a German fashion retailer that wanted to shift consideration for a denim collection among women aged 40 and above. Its usual promotional campaigns ran catalogue ads with photos, which were not moving that metric.

Why the range mattered more than the hero cut

The brand sourced inclusive fashion creators who reflected the diversity of its audience, then produced 22 creator led video assets showing the full denim range across body types and fits. Note what is being varied: not the hook, the body. For an apparel buyer, "will this fit someone shaped like me" is the objection, and one model on one body cannot answer it at any production budget.

The test was a budget split rather than a creative swap. Directing 30% of budget to the creator led partnership ads and 70% to the usual catalogue ads produced a 13% lower cost per purchase and 1.4X incremental return on ad spend, compared with putting 100% into catalogue ads alone.

Rebuild it: cast the objection, then produce the spread

Write down the single sentence your buyer says to themselves before they decline. If that sentence is about fit, size, skin type, room size or hand size, one hero film cannot answer it and no amount of polish will. Cast the range instead:

• Name the objection in the buyer's own words, not in feature language.
• List the variants that would answer it, whether that is body types, skin tones, room sizes or use cases.
• Treat the deliverable as a set, and budget for the set rather than for one film.

Producing twenty two live action cuts is a real budget line, which is exactly the constraint the last three examples here attack from different directions.

3. Boden: built to the placement, not to a master

Boden is a British fashion brand that wanted awareness for a spring and summer collection in the US, and ran a four day burst rather than a long flight. Meta describes it as an intensive four day blast.

Vertical, sound on, safe zone

Boden built the campaign exclusively with video ads featuring creators, following an explicit three-rule checklist:

• Build in the full screen, vertical 9:16 aspect ratio.
• Ensure sound is on.
• Keep the main marketing messages within the safe zone.

The creators showed a range of outfits and highlighted comfort, style and the brand's British heritage. Meta reports a 32.4-point lift in purchase intent.

The interesting part is the third rule. The safe zone exists because interface elements sit on top of your video, and a message placed where a caption bar or a profile row will land is a message nobody reads. Most brands discover this after the campaign, from a screenshot.

Rebuild it: build for one placement, then check the overlays

Pick the placement you actually want to win and master for it, rather than exporting a landscape file and letting the crop fall where it may. Three checks before you upload:

• Master, do not crop. Build to the placement's recommended ratio from the edit, not by trimming a 16:9 file.
• Overlay the interface. Put the app's caption bar, profile row and buttons on top of your final frame in a mockup.
• Move the message up. If your only legible text sits in the bottom third, you have written it into the part of the screen the app owns.

The same discipline applies when you build the campaign itself, which we cover in how to make Instagram ads.

4. BYD: understood in the first seconds, with text

BYD ran creator led partnership ads to build awareness for new electric vehicle models in the UK. Creators shot footage of vehicle interiors and exteriors, and some shot point of view videos using Ray-Ban Meta AI glasses for additional digital first content.

Comprehension comes before persuasion

One line in BYD's creative checklist is worth quoting: ads should be easy to understand from the very first seconds with text overlays. Not "hook them". Understand. It is a lower and far more useful bar, because an ad nobody has parsed cannot persuade anybody, and on an awareness buy the headline underneath will not be there to explain it.

Meta reports a 19.7-point lift in awareness for the SEAL U DM-i model and a 14% increase in Google search driven by Meta, which is the more telling number: people who understood what they saw went and looked it up by name.

Rebuild it: write the overlay before you write the script

Draft the four or five words that will sit on frame one before you write a line of dialogue, and test them this way:

• Name it, do not tease it. A stranger should know the product category from the overlay alone.
• Read it at arm's length, muted. That is the real viewing condition, not your edit suite.
• Assume no headline. On an awareness buy in Feed, the copy underneath may not render at all.

Our guide to ad hooks covers the writing itself, against a hard rule: legibility first, cleverness second.

5. Rexona Australia: a joke only locals get

Rexona wanted a launch campaign that would feel unmistakably Australian. Rather than localising a global master with a voiceover swap, the team generated the visuals.

Local specificity as the hook

The videos reimagined eight famous Australian landmarks, including Parliament House and the Barossa Valley, and used cheeky humour and local slang to connect with audiences across every state and territory. Meta reports a 6-point lift in ad recall for the CGI campaign, and a 37% higher lift in ad recall than the brand's usual campaign.

A landmark your audience drives past is a hook no competitor running a regional master can match, and it costs nothing in media. It costs in production, which is exactly why the team generated the shots rather than travelling to eight locations.

Rebuild it: swap the setting, not the script

Take an ad that already works and change only its world:

• Keep the script. The argument travels; the scenery does not.
• Change the setting to something locals recognise. A landmark, a street, a shop interior, a weather condition.
• Change the words locals actually use. Slang beats translation, and a translated pun is neither.

Generate two or three localised backdrops rather than committing to one global master. It is also the cheapest honest way to find out whether flat performance is a creative problem or a relevance problem.

6. HUGO BOSS: creative iterated by generation, not by reshoot

HUGO BOSS Middle East promoted a winter holiday capsule collection and wanted a diverse set of ads with a premium feel, produced at the speed a seasonal window allows.

Product images, brand guidelines, prompts

The creators used AI image and video generation tools to iterate on creative by inputting product images, brand guidelines, visual direction and written prompts. That produced multiple ad variations, which they then refined for the placement. Meta reports a 63% incremental lift in ad recall and a 5X incremental lift in add to cart events versus the usual campaign alone.

Read the input list again, because it is the actual instruction: product images, brand guidelines, visual direction, written prompts. That is a brief, not a prompt. The generation step did not replace the creative decision, it replaced the reshoot.

Rebuild it: brief the generator like you would brief a creator

Assemble the same four inputs before you touch a tool:

• Product images. Clean, well lit, on a plain background.
• Brand guidelines. Written down rather than assumed, including what the brand will not do.
• Visual direction. One sentence describing the world the ad lives in.
• The script. Written for a person speaking, not for a page.

Then generate a set and cut the weak ones, which is the workflow we walk through in how to make UGC ads with AI. The failure mode is not the model, it is briefing it with one line and hoping.

7. GUMAC: a video ad that ends in a conversation

GUMAC is a Vietnamese fashion brand and an early adopter of livestream commerce. It streams twice daily on average and relies on Facebook for 60% to 70% of its online orders.

Feed to Messenger, with no landing page

Here the video is not the whole ad, it is the doorway. After viewing a live shopping ad in Facebook Feed and tapping Send message, shoppers joined an automated Messenger conversation where they could keep watching, browse an overlay catalogue and check out inside the thread. Optimised for purchases rather than views, GUMAC reports 6.8X higher return on ad spend and a 5.2X lower cost per purchase.

The creative lesson is about the ask, not the film. Every other example here sends attention somewhere else and hopes it survives the trip. This one keeps the entire transaction inside the surface where the attention already was.

Rebuild it: match the destination to the moment

Before you brief the video, decide where the viewer lands and how warm they actually are.

• Cold viewer, three seconds in. Not ready for a checkout page. Ask for the next view, not the card.
• Warm viewer who watched a demo. Does not need another landing page repeating it. Ask for the purchase.
• Product with an open question. Sizing, compatibility, delivery. A conversation resolves it better than a page, so build the ad to hand off rather than to close.

Judging that handoff on clicks alone is how good ads get killed, which we unpack in does a high CTR mean good ads.

How Novoads solves the rebuild problem

Look at what the seven rebuild steps have in common: almost every one asks you to produce a set rather than a film. A range of bodies, a localised backdrop per market, several overlays, a variant per placement crop. That is a production problem long before it is a media problem, and it is the reason most teams keep running the one ad they could afford.

Novoads is an AI UGC video ad generator built for exactly that step. You upload a product image or write a script, pick an AI actor, and get an ad ready video with voice and captions, then repeat it for each variant the teardown asked for. Video runs on frontier engines including Seedance 2.0, Kling v3 Pro and Google Veo 3.1, with per model pricing shown before you render, and output in vertical, square and horizontal so you can master per placement rather than crop and hope. You can try it on a product image you already have and see whether the spread is worth running. Access is $1 for 3 days, then $49 per month. Cancel any time.

Copy the mechanic, not the frame

Every campaign above is already old. The creative is spent, the flights have ended, and the aspect ratio Meta recommended when they were built has moved once already. What survives is the lever each one pulled: the sending handle, the casting range, the placement envelope, the first-second overlay, the local landmark, the generated variant, the destination of the tap. Those transfer to your product, your country and your budget. The frames never do, which is why a folder of screenshots ages into nothing and a list of mechanics turns into next quarter's test plan.