TikTok Video Ad Specs and What Actually Converts

Your ad enters the For You feed in the same slot as a video someone's friend posted twenty minutes ago. There's no banner, no border, no "sponsored" frame that gives the viewer permission to look away. That's the whole problem with TikTok ads, and also the whole opportunity. The specs are trivial. The hard part is not getting skipped in the first second.

This guide covers both: the exact technical requirements so nothing gets rejected or letterboxed, and the creative rules that decide whether a paid view turns into a click. The specs are the easy 20%. The rest is where the money is.

The specs that actually matter in 2026

TikTok will accept a surprisingly wide range of files, but "accepted" and "served well" are different things. Here are the values that keep your ad full-screen, fast-loading, and out of the rejection queue.

Aspect ratio: 9:16 vertical. TikTok supports 1:1 and 16:9, but anything other than 9:16 gets letterboxed in-feed and surrenders the parts of the screen the eye lands on first. Build vertical and crop down for other channels, not the reverse.
Resolution: 1080x1920. Upload at this, not above. Higher source resolution gets re-compressed anyway and buys you nothing.
Length: 9 to 21 seconds is the practical sweet spot for paid. The platform allows much longer, but cost-per-result usually climbs past ~25 seconds because completion rate falls off. Tell one thing well.
File: MP4 or MOV, H.264 video, AAC audio. Keep it under ~500MB; most good 15-second ads are well under 50MB.
Frame rate: 30fps is fine and safe. 60fps looks marginally smoother but rarely changes results.
Safe zones: this is the one people miss. Keep text and key visuals out of the bottom ~140px and the right ~120px. The username, caption, CTA button, and the like/comment/share rail all live there and will sit on top of your content. The top ~130px can also get crowded by the system UI.

The safe-zone rule alone fixes the most common avoidable failure: a perfectly good ad with its offer half-covered by the "Shop Now" button.

Captions are not optional

A large share of feed views happen with sound off, at least for the first beat before someone decides to turn it on. If your message only exists in the audio, you've shipped a silent ad to half your audience. Burn captions into the video itself rather than relying on auto-captions, which can be toggled off and render inconsistently. Keep them inside the safe zone, high-contrast, and large enough to read on a phone held at arm's length.

How the scroll actually works against you

The For You algorithm optimizes for watch time and completion, then layers engagement on top. For a paid placement, the early seconds are weighted heavily because that's where the platform learns whether to keep showing your ad cheaply or tax you with a higher cost-per-impression.

The decision the viewer makes is binary and fast: keep watching, or flick up. They make it before they've consciously processed what your product is. So the first second can't be a logo animation, a slow establishing shot, or your brand name fading in. Those are all "this is an ad" signals, and the trained TikTok reflex is to skip anything that smells like one.

Three things buy you the next two seconds:

Motion on frame one. A static opening frame reads as a paused video and gets scrolled. Start mid-gesture, mid-walk, mid-pour.
A face or a hand, early. Human presence holds attention longer than product-on-white. Even a hand holding the product beats the product alone.
A spoken or captioned claim that creates a small open loop. Not "Introducing X." Something the viewer wants resolved: "I stopped paying for three tools after I found this."

A hook framework you can reuse

Most underperforming TikTok ads fail in the first three seconds, so that's where to spend your creative effort. Here's a structure that works across products. Write to it, then strip anything that doesn't earn its place.

Seconds 0-1 — Pattern interrupt. Visual motion plus a line that doesn't sound like marketing. Either a problem stated as the viewer would say it, or a result stated bluntly. Examples: "This took two minutes." / "My ads were costing me $40 a click. Here's what changed."
Seconds 1-4 — Stakes. Why this matters to them. Name the pain or the desire concretely. Show, don't list.
Seconds 4-10 — Demonstration. The product doing the thing. Screen recording, before/after, a hand using it. This is the proof, and it must be specific, not a montage of vibes.
Seconds 10-15 — Payoff + CTA. The outcome, then one clear instruction. "Link's in the profile." One CTA. Two CTAs split intent and convert worse than one.

A worked example for a hypothetical budgeting app:

0-1: (phone screen, finger scrolling a transaction list) "I found $200 a month I didn't know I was spending."
1-4: "It was all subscriptions I forgot about — three streaming services, a gym I never go to."
4-10: (screen recording) "This flags every recurring charge the day it hits, so I cancel before it renews."
10-15: "Cut my bills by a fifth in a month. App's linked in my profile."

Notice it never says "in today's market" or "the best app for budgeting." It states one specific result and shows the mechanism.

Volume beats polish — why variants win

The biggest mindset gap between agencies and founders is here. A single beautiful ad is a bet on one hook landing. The algorithm and the audience are noisy enough that you can't predict which hook lands. The reliable move is to ship many variants of the same core message and let spend find the winner.

Concretely, for one product you want to test:

3-5 hooks — the same demonstration with different opening lines.
2-3 framings — problem-first vs. result-first vs. comparison.
Multiple thumbnails / first frames — TikTok shows your video, but the first frame still matters for the half-second before autoplay.

That's easily a dozen creatives for a single offer. Producing those by hand or by agency is slow and expensive, which is why most founders ship one ad, watch it underperform, and conclude "TikTok doesn't work for us." The medium didn't fail. The test had a sample size of one.

The discipline that matters: change one variable per variant so you can read the result. If you swap the hook, the music, and the CTA all at once, a winner tells you nothing you can reuse.

Pre-flight checklist before you spend a dollar

Run every creative through this before it goes live. It catches the boring mistakes that quietly cap your performance.

9:16, 1080x1920, MP4/H.264. No letterboxing in the in-app preview.
Captions burned in and readable with sound off.
Nothing important in the bottom ~140px or right ~120px. Preview with the UI overlay, not in your editor.
Motion in frame one. No static opener, no slow logo fade.
Hook contains a concrete claim or open loop in the first second — not your brand name.
One CTA, stated once, clearly.
Length 9-21 seconds. If it's longer, cut the slowest five seconds.
One variable changed versus the other variants in the set.
Sounds and looks native — would this survive in an organic feed, or does it announce itself as an ad in the first beat?

Honest trade-offs

None of this is magic. A few things worth being straight about.

TikTok rewards a native look, which fights brand consistency. The ads that perform often look less polished than your brand guidelines want. You'll have to accept some controlled scrappiness, or pay for it in CPM.

Creative fatigue is real and fast. A winning ad on TikTok burns out faster than on most channels because the same users see it repeatedly in a high-velocity feed. Plan to refresh creative on a rolling basis, not once a quarter.

Volume creates a production problem. The strategy that works — many variants, frequent refreshes, one variable at a time — is exactly the strategy that's painful to execute if every video is a manual edit or a contractor invoice. That bottleneck is the real reason most small advertisers under-test.

FAQ

What's the best length for a TikTok video ad?

9 to 21 seconds for paid placements. Long enough to hook, demonstrate, and ask for the click; short enough to keep completion rate high. If yours runs longer, the fix is almost always cutting the middle, not the ends.

Do TikTok ads need to be vertical?

Effectively yes. 9:16 at 1080x1920 fills the screen and uses the zones where attention lands. Square and landscape are technically accepted but get letterboxed in-feed, which costs you both screen real estate and the native look the algorithm rewards.

How many ad variants should I test at once?

For a single offer, aim for a set of 5 to 12 creatives that share one demonstration but vary the hook and framing, changing one variable per version. Fewer than that and you're guessing; the goal is to let spend surface the winner rather than betting on your own taste.

Producing a dozen captioned 9:16 variants by hand is the part that stops most founders from testing properly. Aitachyon takes a website URL and returns a finished, captioned video ad in about two minutes — three script variants from one brand scrape, exported in 9:16, 16:9, or 1:1 for TikTok, Reels, Shorts, Meta, and LinkedIn. Plans start at $29/mo with a 14-day money-back guarantee, which is roughly the cost of one freelance edit for a month of variants to test against.