2026-05-16
By Vadym · Generated with Boba, curated by me
Think about the last ten screenshots on your phone. Not the memes or the maps — the ones you actually took for a reason. There's probably a banking app showing a balance or a transaction. A password manager with a credential half-visible. A conversation with someone's name and face and words they sent you privately. A work document, a medical result, a boarding pass, a credit card statement. Screenshots are how we grab information in the moment, and what we grab is almost always personal.
That's not a problem until you try to make those screenshots look good.
The workflow for sharing a polished screenshot is well-established: crop it, add a device frame, throw a gradient behind it, maybe blur something sensitive. The tools that do this are mostly web-based or cloud-connected. You open an app or a website, upload the image, apply the effects, download the result. It's fast and the output looks professional.
The part that doesn't get discussed is what happens to the image in between.
When you upload a screenshot to any cloud-based editor, that image travels over a network to a server you don't control, gets processed by software you can't inspect, and sits in storage for an indeterminate amount of time under retention policies you probably haven't read. For a screenshot of a meme, this is fine. For a screenshot of your bank account, your private messages, or a document from work — this is a data handling decision you probably didn't consciously make.
Most people don't think of screenshot editing as a privacy surface. That's exactly why it is one.
Apple's Vision framework makes a different approach possible. Vision does optical character recognition, object detection, and document analysis on-device — on the phone's own processor, with no network call required. It's the same technology that reads text in photos in your camera roll, that recognizes handwriting in Notes, that powers Live Text across iOS.
Applied to screenshots, this means you can detect and classify sensitive text without sending the image anywhere. Recognizing that a sequence of digits matches a credit card format, identifying fields that look like passwords or account numbers, finding names and phone numbers and email addresses — all of that can happen locally, in milliseconds, before the image ever goes anywhere.
This is the architecture Snapdraft is built on. The image stays on the device. The processing happens on the device. There is no backend, no API, no upload. The sensitive content detection and blurring runs entirely through Vision's on-device pipeline. When you tap to share the finished image, you're sharing a file that was generated and modified entirely on your phone.
The practical workflow is straightforward: bring in a screenshot, tap to apply the device frame and background, and Snapdraft runs Vision across the image to identify potentially sensitive text regions. You see what it found and confirm which areas to blur. The blur is applied. The result is a clean, polished screenshot where the information you didn't intend to share has been removed — not uploaded, not processed somewhere else, removed locally before the image goes anywhere.
The design side of this is also real. Device frames and gradient backgrounds are standard visual polish for sharing app screenshots, and doing both in one step without context-switching to another tool is genuinely useful. But the design output is secondary to the data handling model. The reason Snapdraft exists as a native iOS app rather than a web tool is specifically so the processing doesn't require a server.
Security-first design tends to get framed as a constraint — a set of things you can't do, features you can't build, convenience you sacrifice. The on-device approach in Snapdraft is the opposite of that. It's more capable in some ways than cloud approaches: no upload wait, no size limits, no server-side policy about what gets retained. The constraint is architectural, not practical. The user experience is faster.
The broader pattern worth recognizing: a lot of tools handle sensitive data the cloud way by default because building the cloud way is easier, not because it's better for the user. On-device processing in this context isn't a tradeoff — it's a stricter guarantee with no practical downside. The image never leaves the phone because it doesn't need to.
That's the design principle Snapdraft is built on. It shouldn't be unusual.
Snapdraft is free, no ads, no accounts. snapdraft.suprun.ca →