App Store and Play Store Video Assets

Most app teams treat store video as a launch-week chore. Someone screen-records the app on a Tuesday, a designer adds a title card, and that 22-second file sits untouched for eleven months while the product ships four redesigns. Meanwhile the store listing is doing more conversion work than the paid channel feeding it — a shopper who taps through from an ad watches the preview before they decide, and a stale preview showing a UI that no longer exists quietly kills installs you already paid for.

The fix isn't a bigger production budget. It's treating store video assets as a rendering problem: a short script, a set of source clips, and an export pipeline you can rerun in an afternoon whenever the app changes. Apple lets you attach up to three app previews per localization; Google Play takes a single promo video linked from YouTube. That's a small enough surface that a two-person team can keep it genuinely current.

This is the workflow I use for both stores: what each one actually accepts, how to build the product-demo-video content around real screen capture instead of stock footage, and how to keep the whole thing cheap enough to redo every release.

What each store actually accepts

The two stores are different enough that a single master file rarely satisfies both. Build once, export twice.

	App Store (iOS)	Google Play
Slot count	Up to 3 previews per localization	1 promo video
Hosting	Uploaded to App Store Connect	Linked from YouTube
Length	15–30 seconds	30 seconds to 2 minutes recommended
Autoplay	Yes, muted, on the product page	No — user taps play
Content rule	Must be captured from the app itself	Marketing footage allowed
Orientation	Portrait or landscape, must match app	Landscape preferred

Two consequences fall out of that table. First, because iOS previews autoplay muted, every meaningful beat has to survive with no sound — burned-in captions aren't decoration there, they're the whole narration track. Second, because Apple requires genuine device capture, you cannot fake an iOS preview out of generated footage. What AI can do is everything around the capture: the hook card, the context shots, the voiceover, the localized variants, and the Play promo that sits on YouTube. Screen recording is the raw material; generation is the packaging.

The 24-second structure that converts

Store previews fail in a predictable way: they open on a splash screen, then a login, then a settings menu. By second six the viewer is gone. Structure it the way you would a paid social ad, because that's the attention environment you're in.

• 0:00–0:03 — the problem, stated visually. No logo. Show the messy before-state your app removes. This is the only place a generated shot earns its keep in an iOS preview: use it as a title card overlay on the first capture frame rather than a separate scene.
• 0:03–0:10 — the core action. One task, done fast, in the real UI. Not three features. The one thing a new user does in their first session.
• 0:10–0:18 — the payoff. The result screen. Whatever number, chart, saved time, or finished object justifies the download.
• 0:18–0:24 — proof and prompt. A rating, a user count if you genuinely have one, and a clear next step.

Caption every beat: four to six words per card, high contrast, positioned above the bottom third so the store's UI chrome doesn't overlap them. The ideas in brand video hooks: first 3 seconds transfer almost unchanged here, because both are muted-autoplay environments where the first frame does the selling.

Building the capture and the packaging

Here's the concrete loop. Record first, generate second, assemble third.

Record. Use the device recorder on a clean install with realistic seeded data — not test@test.com and three lorem ipsum rows. Capture at native resolution, in the orientation your app supports, and record more than you need: a 40-second continuous run of the core flow gives you room to cut.

Generate the packaging. You need a hook card, sometimes a context shot for the Play promo, a voiceover, and caption styling. On Versely that's a text-to-image or reference-to-video pass for the context shot, text-to-speech for the narration, and auto-timed captions over the assembled cut. The Play version can legitimately open on a generated lifestyle scene — someone using a phone at a kitchen counter, a warehouse floor, a job site — before it cuts to the recording, and that opener is often the difference between a promo that looks like a tutorial and one that looks like an ad.

Assemble. Merge capture and generated segments, add overlays and auto-captions, export. Keep two masters: a portrait 24-second cut for App Store previews and a landscape 45–60 second cut for the Play promo.

For the generated portions, consistency across releases matters more than model choice. Pick one image model and one video model and stay on them so version 3.2's promo looks like version 3.0's. If your opener features a recurring character or your product in hand, a reference-to-video model like Seedance 2.0 Fast Reference-to-Video will hold that likeness across shots far better than re-prompting from scratch.

Localization without reshooting

This is where the effort math flips in favor of AI. Apple lets you upload different previews per localization, but reshooting a voiceover in seven languages used to be a vendor engagement.

The practical approach: keep the visual master fixed and swap only the audio and caption layers. Generate the voiceover per language with TTS, or dub the original narration to keep the same voice identity across markets, then re-run auto-captions against the new track. A fully correct preview also needs localized UI in the capture, which means recording the flow once per language on a device set to that locale — tedious, but a fifteen-minute job rather than a shoot.

A reasonable tier system for a small team:

• Tier 1 (top 3 markets by revenue): localized capture, voiceover, and captions.
• Tier 2: English capture, localized captions only.
• Tier 3: English master, no change.

Keeping it current: the release-cycle trigger

The point of building this as a pipeline is that it survives contact with a shipping roadmap. Tie the refresh to a trigger, not a calendar:

• UI change to the core flow shown in the preview → re-record and re-export. Non-negotiable; a preview showing dead UI is worse than no preview.
• New headline feature → new hook card and caption copy, same capture where possible.
• Seasonal or campaign push → new Play promo only; leave the iOS previews alone.
• Quarterly → review store conversion rate and swap the first preview if it's flat.

Because the generated layers are prompts and scripts rather than files on a designer's laptop, a refresh is an hour of work. Save it as a reusable workflow so the next run is a parameter change — the same reasoning behind treating app install ad videos as templated output rather than one-off creative.

What this costs and what it doesn't fix

Versely bills in credits, so a full asset refresh — a handful of generated shots, one voiceover, captions, and exports — lands in the range of a couple of standard video generations rather than an agency invoice. Check current rates on pricing; free daily credits are enough to test the structure before you commit a release cycle to it.

What it doesn't fix: a preview cannot rescue a listing with a weak icon, a generic first screenshot, or a title nobody searches for. Store conversion is a stack, and video is one layer. It also won't help if your core flow genuinely takes ninety seconds to demonstrate — that's a product problem wearing a marketing costume.

FAQ

Can I use AI-generated footage in an App Store preview?

Not for the app demonstration itself — Apple requires app previews to be captured from the app running on a device. You can use generated content for title cards, overlays, and voiceover, and you can use it freely in the Google Play promo video, which permits general marketing footage.

How long should a Play Store promo video be?

Aim for 30 to 60 seconds. Play doesn't autoplay the video, so a viewer who taps play has already opted in and will tolerate more explanation than an App Store preview viewer. Still front-load the payoff in the first ten seconds.

Do I need separate videos for phone and tablet?

Apple asks for previews matching each device family you support, but the same master cut usually works if your capture is at the right resolution and orientation. Only produce a genuinely different tablet cut if your tablet UI is substantially different from your phone UI.

How often should store video assets be updated?

Tie it to UI changes rather than a schedule. If the screens in your preview no longer match the shipping build, refresh immediately. Otherwise, a quarterly review of store conversion rate is enough to decide whether to test a new opener.

Can captions replace a voiceover for store video?

For App Store previews, yes — they autoplay muted, so captions are doing the work regardless. For the Play promo, a voiceover plus captions performs better because the viewer has chosen to play it with sound.

Ready to rebuild your store assets? Start with a capture of your core flow, then use the AI product video generator to build the hook, voiceover, and caption layers around it — and save the result as a workflow you can rerun every release.