A Shopify app-store listing is screenshot UI, not a theme tile

Shopify theme listings (already a post) are storefront stills. An app listing is the admin screenshot row: settings, install button, nav type. /for/shopify-stores is the merchant. Qwen or Recraft holds the chrome.

An app that sells is a screen: settings grid, install button, nav labels a merchant will proof against the live admin. Motion is optional. A constructed shopper walking a cart is a store ad. It is not the listing.

Theme tiles already have a post

A Shopify theme listing is a storefront still, not a store ad already owns the homepage: header, nav words, a price that looks like money, a product grid a merchant will recognise as this theme. Recraft or Seedream holds that UI. Do not rewrite it as an app tile.

An app listing is a different chrome. Admin, not storefront. Settings, not a product card. Install, not Add to cart. Cropping a theme homepage into an app screenshot is how a buyer lands on the real App Store and sees a cousin product. The two objects share a platform. They do not share a still.

Do not generate a “busy boutique on a Saturday” and crop it into either tile. Generated shoppers are a likeness problem and fake social proof. Empty UI is honest. Mixing the two objects is how an install button grows a fake PDP.

The listing is the admin screenshot row

Digital tools have no unboxing. The fear is a thin file and a fake panel. Capture the real app you would ship: setting names matching the store copy, an install button that is the button you will upload, nav type that matches the live admin. Crop chrome that is not the SKU. Crop other merchants’ store names if a test shop leaked one. The still is a document. Treat it like one.

Do not prompt a video model to “show a premium Shopify app with cinematic productivity energy.” It will invent a cousin admin, a cousin install glyph, a cousin nav. A merchant who lands on the real listing and sees a different settings set is not looking at creative. They are looking at a lie.

If you do not have a screenshot yet, you still do not invent a shopper. You typeset a mock of the admin. Quote every string a buyer will proof: Save, Install, the nav labels as they will appear. A photoreal founder pointing at a MacBook is a presenter. The SKU is the screenshot row. Wrong chrome is a cousin product.

Qwen or Recraft holds the chrome

Two stills rows, two meters. Pick the one the screenshot needs.

Qwen Image 3 Text to Image is the stills row when the letters have to read. Catalog facts: Qwen’s third-generation image model, strong prompt adherence, complex text rendering in Chinese and English, intelligent prompt rewriting on by default, 4 credits, 2K max. No audio. No duration. Aspects include 1:1, 16:9, 9:16, and 21:9.

Write the exact strings. If a setting name or an install label must stay verbatim, say so in the prompt and inspect the still. Rewriting is a help for scene language, not a licence to paraphrase Install app into Get it. Set 1K while the words are in draft. Set 2K when the admin is the admin. There is no 4K chip on this row.

Recraft V4 is Recraft’s latest text-to-image model with design quality and typography support. Five credits, flat per call. audio is null. No duration. Aspects include 1:1, 16:9, 9:16, 4:3, 3:4, plus layout ratios such as 14:10. Pick the ratio of the listing. Generating 1:1 and cropping later is how the nav hierarchy dies. Quote "Install", "Settings", the nav type you actually show. Proof at 100%. A wrong character in the nav is a different app. Treat the output as a raster still.

Generate an image from text is the named agent job: describe the admin screenshot, the agent asks style, aspect, and count before it spends, then picks a catalog image row — or the one you name. Cost is priced per model, shown before you confirm. Name Qwen or Recraft.

Text to image is the tool-shaped door for the same still. A video model is a bad typography engine. Approve the characters. Then consider motion. If the printed chrome is already sharp in the screenshot, you may not need a generate at all.

/for/shopify-stores is the merchant, not this author

Shopify Stores is the nearby audience and a different page. That playbook is product-page-adjacent video: the clip that should live on the PDP, UGC-style reviews, comparison ads against the category default. The merchant is the user. This listing is the app author. Do not rewrite /for as an App Store screenshot. A candle unboxing is not nav type.

Do not copy Shopify Stores film ideas onto the tile: unboxing, a size demo, a customer testimonial. Those are merchant jobs. The listing has to survive a zoom on the install button. Lock the admin on Qwen Image 3 or Recraft V4; proof every label; export the tile. Animate only if the brief truly needs a hover, and then from that still.

Do not Veo a fake shopper. A talking cinematic row will still draw a person and letterforms. Those letterforms will not be your nav. Empty UI. Readable settings. Prefer the still. If a reviewer would not say “that is this app’s settings row,” you made a store ad.

FAQ

Should I generate the App Store tile as a Veo clip of someone installing?

No. A fake shopper is a likeness problem and a bad typesetter. Qwen Image 3 (4 credits, 2K) or Recraft V4 (5 credits) holds the admin as a still.

Is /for/shopify-stores the page for app authors?

No. That page is the merchant: PDP-adjacent video, UGC-style reviews. An app-store listing is admin screenshot plus settings type. Do not copy the merchant playbook onto the tile. The theme listing post is storefront chrome, not this admin row.

Qwen Image 3 or Recraft V4 for a dense settings grid?

Qwen Image 3 is 4 credits, 2K, complex text in Chinese and English. Recraft V4 is 5 credits, typography, design-quality stills, layout aspects including 14:10. Proof the nav either way. Neither row is a video.

Can the agent just “make a Shopify app video”?

Generate an image from text is a stills job. Name Qwen or Recraft. A video brief will route you toward a talking shopper. Approve the still. Then stop, unless a later image-to-video pass is actually required.