Windows app
Grab a region, any window, or your entire screen — then ask it what it means, read its text, fix its error, or turn it into a table. Your captures stay on your machine.
20 AI credits a month on the free plan. No card, no API key to set up.
Every AI feature in ScreenFeel — every lens, every follow-up question, every automatic title — is answered by a model our inference provider runs on its own hardware, under zero data retention. Your capture is read for your answer and is then gone: never handed to a consumer AI service, never written to a log or a dataset, never turned into training data. There is no cheaper second path we fall back to, and no setting that turns this off.
Inference providers also sell a cheaper anonymous tier: it hides who you are, but an outside AI lab still reads what you sent. We do not use it, for anything. We call only open-weight models the provider hosts itself, so no third-party lab is ever in the path.
Your image and its answer exist for the length of one request and are not retained afterwards. Nothing is stored that could leak later, and nothing survives to be trained on. Your account does not travel with the request either — the model host sees our operator key, never you.
Which models we call is configuration, so it is verified rather than trusted: a check we run before every deploy reads the provider's live catalog and fails if any model we would call is not marked private.
Sixteen lenses, each one a different, very specific question about the picture.
Works out what this is and does the most useful thing
What it is, why it matters, what is off
Verbatim transcription, layout preserved
Charts and tables become numbers you can paste
Every word, into your language
Reads the trace and names the likely cause
A designer's eye on this interface
The substance of a long page in a few lines
Code out of a screenshot, ready to run
Works through the maths on screen
Spots anything you would not want to share
How this screen reads to everyone else on the call
Accessible description, ready to paste
Turns notes and threads into a to-do list
Writes the message you were about to send about this
What to say about this screen, out loud
None of this makes a network call. It is the part of the app you would use even with the AI switched off.
Region, window or screen — in any app, one keystroke each
Annotate, redact and export at native resolution
Cursor glow, spotlight boxes, arrows and fading ink over any app
Every capture, searchable, on your machine
Two more lenses take a set of captures from the Library rather than a single picture.
Nobody else sells the same combination: capture anything, keep it on your machine, and send exactly one image — only when you ask — to a model that keeps nothing.
A screen-aware AI assistant shaped around live meetings and interviews.
A desktop application that can see your screen and answer questions about it.
A broad browser AI sidebar with chat, agents, image generation and document tools.
A privacy-first AI assistant from Proton, with zero-access encryption and no training on your data.
The AI assistant built into the Brave browser, proxied so the provider never sees your IP.
Microsoft's screen-aware assistant, free in Edge and Windows 11.
A browser AI sidebar with a large free allowance and a low paid entry price.
A region with a pixel-exact magnifier. Any window, picked from a live grid. Or the whole screen — in any app, because ScreenFeel answers the Print Screen key itself.
For Windows 10 and 11. Capturing, annotating and the library are free without an account — signing in switches on the AI lenses.
Get it for WindowsInstalled and kept up to date by the Microsoft Store.
We run the AI so you never have to hold an API key, and every model we call is a private one — hosted by our inference provider itself, under zero data retention. Your screen is not handed to a consumer AI service, and it never becomes training data.
Every screenshot, thumbnail, annotation and chat sits in the app's own storage on your device. There is no upload, and no gallery of yours on our side.
Press a lens and that single image is sent, read, and answered as text. Turn automatic naming off and nothing leaves at all until you press.
Every lens runs on a private model under zero data retention. Your capture is not stored after the answer, never used to train anything, and carries no account of yours — nobody on the other end has any idea who you are.
ScreenFeel flags key-shaped and card-shaped strings in anything it has already read, on your machine, before you paste a picture somewhere public.
Lenses are weighted, because they differ in cost by a factor of twenty-five. Most work is a single credit, and nothing is locked behind a plan.
Capture, annotate and search your library forever. 20 AI credits every month, and nothing to configure.
Get it for WindowsCookiesThis site remembers the language you pick and, only if you allow it, counts anonymous page views. Nothing here follows you between sites, and there are no advertising cookies. Read the privacy policy.