Media
Media keeps your images, video, and visual references in the same workspace as everything else, so the picture sits right next to the thinking.
A lot of what you want to remember is not text: a screenshot of a bug, a whiteboard photo, a reference image, a short clip. Media gives those a home inside your workspace instead of scattered across a downloads folder, where they can be searched, arranged, and linked like any other item.
When to reach for media
- collecting reference images or video for a project
- keeping a screenshot or a diagram right beside the note that explains it
- holding source material you will write about later
- annotating an image and pointing notes back at it
What it handles
Media takes in images and video and keeps them as real files in your workspace. Images cover PNG, JPEG, WebP, and GIF. Video covers MP4, WebM, MOV, MKV, AVI, and M4V. For images, Undra can also pull text out of the picture with OCR (described below), so even visual material becomes searchable.
Add media to your workspace
The simplest way to get a file in is to drag it onto the Media view. Drag one or more images or clips from your desktop, a folder, or another app straight onto the Media grid (or onto a specific folder in the sidebar). Undra copies the file into your workspace, so the original on disk stays where it was and the workspace now has its own copy. Drop onto a folder and the imported items land in that folder; drop onto the grid and they go in unsorted.
Media also fills up as a side effect of normal work: images you reference inline in a note or pull onto a canvas become part of the workspace the same way. The point is that you do not manage a separate upload step. You put the picture where you are working, and it is now an item you can search, tag, and link.
Anything Undra cannot handle as an image or video (the formats listed above) will not import as media. If you need to keep a file of another type next to your work, attach it to the relevant note or plan instead.
Keep it next to the work
Media earns its place when it stops being a separate gallery and becomes part of the work:
- Reference it from a note. Drop an image straight into your Markdown so it renders inline next to the writing.
- Arrange it on a canvas. Pull images and clips onto a canvas and lay them out as a moodboard or a visual map.
- Attach it to a project. Keep the relevant screenshots beside the plan or note they belong to, so the context travels with the work.
Annotate an image
You do not need to leave Undra to mark up an image. Right-click an image in Media and choose Open in image editor to draw on it: freehand strokes, shapes, arrows, and text, with undo, redo, and a colour picker. Your annotations save with the image and come back when you reopen it, and you can export the marked-up version as a PNG.
Annotations are non-destructive. They are stored as a separate overlay layered on top of the picture, not painted into the original file, so the source image is never overwritten. You can change or remove your markup at any time and the underlying image is still pristine. The PNG export is a fresh, flattened copy with the annotations baked in; it does not replace the original either.
This is for annotating, not photo editing. There is no crop or filters, just a clean way to point at the thing that matters.
Find it later
Because media lives in the workspace and not in some external bucket, it shows up where you look for everything else. Search reaches it by name and by tag, and any text Undra pulled out of an image with OCR is searchable too, so you can find the screenshot by what it said rather than by remembering its filename.
That OCR runs locally on your machine, on images only (not video), and nothing leaves the box. It happens automatically in the background when you open the Media view: Undra works through your images, reads the text it finds, and stores it so search can use it. To keep the index clean it only keeps text it is reasonably confident about, so a photo with no real writing in it does not pollute your results. English, Simplified Chinese, Japanese, and Korean are read out of the box, and you can opt in to more languages from the AI settings. The first run for a language downloads a small recognition pack and caches it for next time.