Docs

1. 01Open the studio and drop in screenshots straight off your phone. No account needed.
2. 02Dimensions are read locally to identify the device, which fixes the frame, corner radius, Dynamic Island geometry and status-bar height.
3. 03Copy is generated per screen. Edit any of it — the preview updates as you type.
4. 04Export. You are asked to sign in only at this point.

Every canvas Shipshot renders, and the slot that accepts it.

Apple accepts a maximum of 10 screenshots per display size per localisation, and shows the first three on the install sheet. Google Play accepts 8.

This is what the upload box in App Store Connect will actually accept, per slot, in both orientations. Anything else bounces.

This is the single most common cause of a rejected upload. App Store Connect validates a file against the slot you drop it into, not the device you captured on. A perfectly valid 1320 × 2868 file from an iPhone 16 Pro Max is rejected by the 6.5" slot, which wants 1284 × 2778 or 1242 × 2688.

All four iPhone canvases are within 0.5% of the same aspect ratio, so one capture set legitimately feeds all of them. Shipshot renders every one and names each output folder after its slot, then reads each PNG's header back to confirm its dimensions before you download.

Every text field on the version page, with the limit App Store Connect enforces. Hitting one of these halfway through a submission is the classic way to lose an afternoon.

Your binary can be perfect and your screenshots pixel-exact, and you will still be stuck in Prepare for Submission until these three are done. None of them are code.

Headlines are written by Claude Fable 5 (claude-fable-5) from a description of what is on each screen. The model is instructed to:

• Never invent a feature that is not visible in the capture.
• Keep headlines under 38 characters and state a benefit, not a screen name.
• Name one concrete, visible detail in the subline.
• Wrap exactly one phrase in *asterisks*, which renders in your accent colour.

Without an ANTHROPIC_API_KEY, the API falls back to deterministic copy so the demo never fails on a fresh clone.

Copy generation is exposed directly. Team plans get export endpoints for CI.

POST /api/generate-copy
{
  "appName": "Stock Picks",
  "screen": "Order filled confirmation, BUY 1 SPY @ $747.03",
  "category": "fintech"
}

→ { "headline": "Orders fill *instantly*",
    "sub": "Confirmed at the ask, at the price you were shown",
    "source": "claude-fable-5" }

Pass screens: string[] instead of screen to generate a whole set in one call — cheaper, and the copy varies more across the set.