Send a screenshot to AI without leaving Windows

Capture a region from any app

Press Ctrl+Shift+X in whatever app you are using and drag over part of the screen. The region is attached to the message you are writing in SideNote Pro, beside any text and attachments already there, and the cursor waits in the message box for your question. Nothing is sent, and no new conversation is started. Esc or a right-click cancels and leaves everything as it was.

The shortcut captures the screen as it is, without hiding SideNote Pro first, so a SideNote Pro window that is on screen, such as a floating note, can be part of the region. The capture button in the composer hides SideNote Pro before it captures, as before.

The shortcut can be changed or cleared under Settings › Actions & prompts. While SideNote Pro runs, it takes precedence over the same keys inside other apps, so if an app you use relies on Ctrl+Shift+X, choose another. It does nothing while a response is being written or when five images are already attached, and says why.

Four capture modes

Region capture temporarily hides SideNote Pro, captures the virtual desktop as a visible backdrop, and shows a translucent selection overlay. You drag out the area you want and see a blue selection rectangle; Esc or a right-click cancels. The selected region is attached to your pending request as a PNG.

Active-window capture remembers the foreground window from the summon or capture flow, so it captures what you were looking at rather than the assistant itself, and uses DWM visible-frame bounds where available. Capture works in physical coordinates throughout, so multiple displays and negative virtual-desktop positions behave correctly.

Other ways to attach an image

The image picker

PNG, JPEG or JPG, and WebP.

Drag and drop

onto the assistant window.

Clipboard paste

with Ctrl+V when the clipboard holds a bitmap. Useful with the Windows snipping tools you already use.

The built-in screenshot tool

described above.

Up to five images can be pending per request, and PDF pages read as images take the same five slots. They are previewed locally before sending, can be removed from the composer individually, and are kept in deterministic order so the first image means what you expect. They travel to your provider as base64 data URLs in the OpenAI-compatible multimodal message shape.

An image can be up to 20 MB, 16,384 pixels on a side and 64 million pixels in all, and the images in one message up to 50 MB together. A larger image is refused with a message saying which limit it passed, and what is already in your message stays.

Your model needs image support

Multimodal input is a model capability, not an application one. SideNote Pro can format and send the image correctly to any OpenAI-compatible endpoint, but a text-only model will refuse it or ignore it. If images seem to be dropped, check the model rather than the app. This applies to local providers too: many small local models are text-only.

SideNote Pro does not perform OCR of its own. Whatever text a model can read out of your screenshot, it reads with its own vision capability.

Nothing is watching your screen

Screen awareness has become a common feature in this category, and SideNote Pro does not have it. There is no background capture, no periodic screenshotting, no ambient screen understanding and no monitoring service.

A screenshot exists because you started one, from the composer or with the capture shortcut. A clipboard image is read because you pressed Ctrl+V. And an attached image is transmitted only when you send the message it is attached to. Protected or DRM-restricted surfaces may still come out blank, because Windows blocks their capture regardless of who is asking.