How to Make Static Ads From Your Best Video Frames

Open the video that earned a 30%+ hook rate (or your current best open) and scrub only the first two seconds plus the single clearest proof beat. Export three to five frames where the face, product, or result is large, sharp, and readable with the sound off. Ignore cinematic mid-shots. A static ad has one job: the same thumbstop, frozen, on a phone at arm's length.

The first-frame rule still applies when the image does not move. If the video won because a stain filled the screen, that stain is the static — not a later smile. If it won because of a confession on a close face, export the face with the mouth shape that still looks like speech, plus a frame of the product in hand. Blurry motion frames feel 'energetic' in a timeline and look like mistakes as stills; if the motion is the point, pick the moment just before or after the peak blur. Export at native resolution. Do not screenshot the Ads Manager preview.

Take frames from more than one cell of the 3 × 2 grid if more than one hook worked. Different first frames become different static concepts. Taking five crops of the same smile is one ad. If no video has hooked yet, you can still cut frames for a parallel static test, but do not pretend they inherited a winner. Label the export with the video's hook name so the static keeps the same hypothesis when you read results. A still without a hypothesis is a design file, and it will not teach you anything in 48–72 hours.

Make a 1:1 and a 4:5 (or 9:16 story) from each keeper frame. Keep the face or product in the centre third so Feed UI, the CTA bar, and story chrome cannot cover it. Do not letterbox a 16:9 frame and call it a static. Recrop. The still should feel native to the unit, not like a paused trailer sitting in a black box. If the crop needs a caption to explain the object, pick a closer frame.

On 1:1 Feed, leave breathing room at the bottom for the headline and button Meta will draw. On 9:16, keep text and faces out of the top and bottom UI bands the same way you would for Reels captions. If the interesting part of the video frame is in a corner, it will die on mobile. Zoom until the subject fills most of the canvas — UGC statics that perform usually look closer than a designer would crop a brand ad. Grain and handheld oddness can stay; they are part of why the video worked. Do not 'clean up' a winning frame into a catalogue shot unless you are deliberately testing catalogue.

Check the crop on a phone at arm's length for one second. If you cannot name the product or the problem, recrop or pick another frame. This is the same mute test as video, minus time. A static that needs a paragraph of primary text to make sense is a weak frame wearing copy. Fix the picture first. Keep the original uncropped export so you can recrop later without regenerating the video. If the interesting bit only lives in a corner, it is not a static — it is a still from a wide shot you should not have frozen.

Write six to ten words that match the video's hook or the single proof the frame shows. Place them in a caption-safe band, high contrast, one weight. Do not stack a headline, a subhead, three bullets, and a badge. The frame is the ad; the overlay is a caption for mute. If the picture already shouts, you may need no overlay at all — test both. One line of type, or none — never a poster of claims on a UGC still.

Match the spoken hook when you can. If the video said 'I was ready to return this,' that line on the still is the inherited thumbstop. Do not swap in a brand slogan because static 'should look more designed.' Use the same claim rules as the video: no health or income superlatives you cannot stand behind. If you include price or a deadline, you have changed the job of the ad toward offer-led retargeting — that is valid, but it is a different test cell, so name it that way. Keep logo small or off the first read; a giant logo on a UGC frame turns it back into a banner.

Set type as if it were a burned-in caption: bold sans-serif, large enough to read at arm's length, one or two lines. Avoid placing text over the mouth or the product detail the frame was chosen for. If you need a button-like 'Shop the bundle,' remember Meta will add its own CTA — duplicate buttons look like a mistake. Export a no-text version of every static so you can test overlay versus clean frame. Many UGC stills win without type because the face or the mess already does the hook, and extra copy only makes them look like banners.

Give every still a job: cold thumbstop, social proof, or offer-led retargeting. Cold statics should look like the video open. Warm statics can carry price, bundle, or 'still thinking it over.' Do not run an offer-led still to cold traffic and then blame the frame. Write the job in the file name before upload so the readout still knows what you meant to test.

A practical mix from one winning video: two cold stills from the first two seconds (one face, one proof), one proof still from the demo beat, and one offer still for view-through or site-visitor retargeting. That is four files, not a 40-asset design sprint. Primary text in the ad unit can stay short and consistent with the overlay. Use the same landing page as the parent video unless the static is a genuine offer test. Changing the page and the format at once is how statics get a reputation for being 'unpredictable.'

On Meta, statics still belong in a structured test, not a junk drawer. You can run them in the same prospecting ad set as video if you want the platform to pick format, but then you must read results by format and by concept, not as one blended ROAS. Cleaner: a separate ad set or a clearly named subset for static concepts, same audience, 48–72 hour window, equal chance to reach ~1,000 impressions. Advantage+ will mix them whether you planned to or not — naming is how you still learn. If you cannot say which still was the cold open, you did not run a static test, you ran a collage.

Run the statics for 48–72 hours with the same patience you give video. Keep winners in retargeting and as Feed companions. When the parent video hits frequency 2.5–3.5 on cold traffic, do not assume the still is tired — a new crop or a new overlay from the next hook can extend the asset. Retire the still when its own CTR and CPA die, not when the video file is paused.

Judge cold statics on outbound CTR (aim for a 1%+ cold floor, knowing static CTR is not identical to video) and on CPA, with a glance at whether the image is actually stopping anyone. There is no 3-second hook rate on a still; do not invent one. If a static from a winning video fails cold, it may still work in a 7–14 day viewer or visitor pool where recognition does the thumbstop. That is a different cell. Move it rather than deleting it. Offer stills with a deadline should be paused when the deadline dies; leftover urgency is how accounts look sloppy.

Build a small frame library next to your swipe file: winning first frames, labelled by hook family. Next week's 3 × 2 video batch should produce next week's statics on the same afternoon, not a month later in a design queue. Klip Kanvas can also generate static variants from the same product brief — use that when you need extra crops, but still prefer frames from a video that already hooked when you have them. Inherited thumbstops beat invented layouts. When the parent video later hits frequency 2.5–3.5, recrop from the next hook instead of treating the still as automatically dead.