Product Visuals logo

Product Visuals

Community
Innei
product-visuals

Use when the user wants a product marketing image from real macOS or iOS screenshots — dual-device hero, device-framed mockup, 宣传图, 产品图, 机型边框, privacy-safe Safari or web presentation, App Store screenshots, OG card, changelog card, or to swap the background of an existing device mockup. Also when they run /product-visuals.

Overview

PublisherInnei
RepositorySKILL
Skill nameproduct-visuals
Stars
81
Forks
2
Bundled files
5
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 5 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by Innei on GitHub. Read the source before you install it.

Installation

Install the Product Visuals AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/Innei/SKILL.git /tmp/SKILL
mkdir -p .claude/skills
cp -r /tmp/SKILL/skills/content/product-visuals .claude/skills/product-visuals
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Product Visuals in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Product Visuals on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Product Visuals is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Product Visuals

Compose product marketing stills from real app screenshots inside official Apple device bezels. Background is a separate layer and can be swapped without re-framing the screens.

Capability contract

  • Outcome: a 16:9 (or requested) PNG with iPhone / MacBook frames and pixel-accurate UI
  • Preconditions: iPhone screenshot and/or real app/browser window screenshot; network access to fetch Apple resources and public wallpaper when needed
  • Boundaries: do not invent or redraw UI; do not implement app-store-set / og-card / changelog-card in this version (route only)

Iron rules

  1. Never send product screenshots through an image model. Frames and composite are local code.
  2. Never commit or vendor Apple Design Resource PNGs/DMGs. Fetch them under ~/.cache/product-visuals/.
  3. Clip the screenshot to the interior screen mask, not the hole's bounding box (bbox corners sit outside the rounded device).
  4. Zero RGB on fully transparent bezel pixels and premultiply before LANCZOS resize.
  5. Shadows are padded contact shadows under the chassis. Do not blur the full bezel canvas — that stamps a rectangular plate.
  6. Never use the user's live desktop, personal wallpaper, notifications, bookmarks, or unrelated windows unless they explicitly request it.
  7. For a web product on Mac, screenshot a real local Safari window with its native shadow (screencapture -l <CGWindowID>). --mac accepts that window only; the compositor supplies the wallpaper and Menu Bar and never adds a second window shadow. Do not paste a webpage screenshot into someone else's Safari chrome. Do not redraw Safari.
  8. For iPhone, capture the simulator framebuffer with xcrun simctl io <udid> screenshot. Never screenshot the Simulator.app window. Never crop a screen out of an already-framed mockup.

Recipes

RecipeWhenStatus
dual-device-heroiPhone + Mac window, site hero / 宣传图implement
app-store-setApp Store screenshot setstub — say so, do not fake it
og-card1200×630 share cardstub
changelog-cardcompact update visualstub

Default composition for dual-device-hero: 4800×2700, opaque device group centered on the canvas, MacBook Pro 14 Space Black slightly left of that group, iPhone 17 Pro Silver overlapping lower right.

Workflow

text
[1] Pick recipe from the request (default dual-device-hero)
[2] Collect --phone (simctl framebuffer PNG) and --mac (real Mac window PNG with native shadow)
[3] Export the official Menu Bar SVG from Apple's macOS UI Kit and cache it with scripts/cache-macos-menu-bar.sh
[4] Let compose.py build the Mac screen as wallpaper -> official Menu Bar -> --mac window
[5] bash scripts/fetch-bezels.sh
[6] Derive background from the app's 调性 (see Background)
[7] uv run scripts/compose.py dual-device-hero --phone … --mac … --bg … --out …
[8] Open the PNG. Confirm UI is readable, no corner overflow, no rectangular plates around devices, background reads as the same product
bash
HERE="<path-to-this-skill>"
bash "$HERE/scripts/fetch-bezels.sh"
bash "$HERE/scripts/cache-macos-menu-bar.sh" "$MENU_BAR_SVG"
uv run "$HERE/scripts/compose.py" dual-device-hero \
  --phone "$PHONE" --mac "$MAC" --bg "$BG" --out "$OUT"

Swap background only — same screenshots, new --bg:

bash
uv run "$HERE/scripts/compose.py" dual-device-hero \
  --phone "$PHONE" --mac "$MAC" --bg "$BG" --out "$OUT"

--iphone-color silver|deep-blue|cosmic-orange and --mac-color space-black|silver select cached official bezels.

Background

The plate is the product's 调性, not a stock cinematic studio. Derive it from the screens (and DESIGN.md / tokens if the app repo is known). Do not rebuild device frames to change it.

Yohaku example: 余白, warm parchment #f9f8f5 / dark desk #141414, 梅 #c56473 ≤5% — paper and ink, not blue volumetric cinema.

User saysDo
nothing / defaultname the app's palette + material from the screenshots (and DESIGN.md). image_gen / gemini-image-generation: empty 16:9 plate in that 调性, no devices, no screens, no text; then --bg
cinematic / 更电影generic empty studio only when they asked for it
here's a file / 用这张底--bg that file
omit --bg / generation unavailablecompose.py samples the screenshots into a quiet gradient
换背景 on an existing herokeep the same --phone / --mac; new --bg still from that app's 调性

Generated atmospheres must stay empty. If a model draws a laptop or phone, discard and regenerate.

--bg-dim 0.88 darkens a too-bright plate. Default is 1.0 (unchanged).

macOS web presentation

When the Mac screen presents a website, read references/macos-web-scene.md before composing it. The --mac input is only the real Safari window capture. The compositor builds the privacy-safe desktop scene from separate wallpaper, Menu Bar, and window layers.

iPhone capture

--phone is the device framebuffer, not a photo of Simulator.app.

bash
UDID=$(xcrun simctl list devices booted | awk -F '[()]' '/Booted/{print $2; exit}')
xcrun simctl io "$UDID" screenshot phone.png

Accept: a rectangle at the device's native screenshot size (iPhone 17 Pro: 1206×2622). The status bar and iOS's black island pill are framebuffer pixels; the official bezel supplies the hardware island and camera.

Reject and recapture:

  • Simulator.app window (screencapture, axe of the window, a display screenshot that includes chassis chrome)
  • A crop from an already-framed mockup (camera already in the island, rounded device corners in the PNG)
  • Any size other than the booted device's native screenshot pixels

The workflow's "real window capture" is Mac only.

Inputs

FlagMeaning
--phoneSimulator framebuffer PNG from simctl io screenshot (iPhone 17 Pro: 1206×2622)
--macreal app/Safari window PNG with native window shadow; never a full desktop capture
--mac-wallpaperoptional public wallpaper override; otherwise use the built-in privacy-safe wallpaper layer
--mac-menu-baroptional transparent official Menu Bar PNG override; otherwise use the cached macOS 27 UI Kit layer
--bgoptional background image, any aspect; resized to the canvas
--outdestination PNG

Reject a --mac capture that includes wallpaper, a desktop corner, the display notch, Dock, notifications, or unrelated windows. Recapture only the target window.

Verification

Before claiming done:

  • Open --phone before composing: rectangle at native size, no silver chassis, no camera lens in the island.
  • Open the PNG. Read actual UI text on both screens — it must match the source screenshots.
  • Check iPhone top-left and both bottom corners: no screenshot rectangle leaking past the silver frame.
  • Check the Dynamic Island: one hardware island with the camera. A second black pill beside or below it means --phone was not a framebuffer shot, or the island was not covered by the bezel.
  • Check around both devices: no darker rectangular plate, no second shadow card on the Mac.
  • Check the MacBook display corner against the official bezel: the screen content reaches the official interior mask and has no second inset rounded corner from a desktop screenshot.
  • Check the pair as a group: left/right gaps around the opaque chassis should match. A left-heavy Mac with empty canvas on the right means the layout used PNG origin instead of the opaque union.
  • Background shares the product's palette and material. A blue cinematic void behind a parchment/ink app is wrong.
  • For a web product, confirm the browser is genuine Safari, fills roughly 84–90% of the desktop width, and contains no private browser or desktop data.
  • Confirm Safari has one native rounded border and one native soft shadow. A second hard or concentric arc means the compositor added or reconstructed a shadow; it must not do so.
  • Confirm the Menu Bar is the unmodified official UI Kit export: Apple glyph, outlined typography, status icons, spacing, and date/time must remain coherent.
  • If only the background should change, the framed screens must be identical to the previous export.

Common mistakes

MistakeFix
Image-model restyles the UI into the devicesCode composite only
alpha < 16 inside the screen bbox as the holeUse the interior flood-fill mask
Gaussian-blur the full bezel image for a drop shadowPadded contact shadow under the base
Resize RGBA without premultiplyDirty transparent RGB becomes a gray halo
Check Apple bezels into gitCache only; fetch-bezels.sh
Pin Mac at (150,150) / center the bezel PNG canvasCenter the opaque-union; 14" Mac mockups have ~230px empty top pad
Default to cinematic blue-rose studioRead the screens (and DESIGN.md). Generate or sample from the app
Pass a full desktop screenshot as --macPass only the target window; let the compositor own wallpaper and Menu Bar layers
Hand-draw Safari or paste a site shot into another window's chromeResize the live Safari window, capture that real window, and composite its PNG with alpha
Add a drop shadow around the Safari PNGKeep the native screencapture -l shadow; the compositor adds no window shadow
Repair rounded corners from a desktop screenshotRemove that code path; a wallpaper-first rectangular scene has no baked display corners
Recreate the Menu Bar with local fonts or hand-drawn iconsExport the official component from Apple's macOS UI Kit and cache it with cache-macos-menu-bar.sh
Screenshot of Simulator.app, or a crop from a framed mockupxcrun simctl io <udid> screenshot at native size
Two Dynamic Islands / camera floating next to a black pillRecapture with simctl; do not feed an already-framed PNG as --phone

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Product Visuals AI skill do?

Use when the user wants a product marketing image from real macOS or iOS screenshots — dual-device hero, device-framed mockup, 宣传图, 产品图, 机型边框, privacy-safe Safari or web presentation, App Store screenshots, OG card, changelog card, or to swap the background of an existing device mockup. Also when they run /product-visuals.

Why use Product Visuals on TypingMind?

Because you install it once and use it with any model. Product Visuals is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Product Visuals in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/Innei/SKILL/tree/main/skills/content/product-visuals. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Product Visuals?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Product Visuals?

As many as you like. As long as a model supports skills, you can use Product Visuals with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Product Visuals AI skill free?

It is published on GitHub by Innei. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇