Drop in a screenshot, drag boxes around the parts you are pointing at, and download it. Any colour, with an optional translucent fill that highlights an area without hiding the interface underneath — usually the thing you are pointing at is the thing that has to stay readable. Boxes are stored proportionally, so they land exactly right on the full-resolution export rather than at preview scale, and the stroke thickens with the image so it reads clearly on a 4K capture as well as a small one. PNGs stay PNGs, which matters: re-encoding screen text as JPEG puts visible ringing around every letter.
| Goal | Tool |
|---|---|
| Point something out | This — outlined box |
| Highlight an area | This — translucent fill |
| Hide information | Photo Censor — solid block |
| Remove GPS data | Image Metadata Remover |
No. It is annotated in your browser and never transmitted.
Because converting a screenshot to JPEG puts compression ringing around every letter of screen text, which makes it look degraded. Keeping PNG keeps the text crisp.
No — they are drawn into the image. Undo works while you are editing, but the exported file has no separate layer. Keep the original if you might revise it.
Yes, at full resolution. Boxes are stored as proportions, so the preview size does not affect where they land in the output.
It scales with the image. A fixed two-pixel line is clear on a small screenshot and nearly invisible on a 4K one, so the width follows the resolution.
Use the Photo Censor for that — it has a solid block designed for redaction. A translucent fill here is a highlight, and whatever is underneath is still visible.
Boxes are stored as percentages of the image rather than pixels, so a rectangle drawn on a preview a few hundred pixels wide lands exactly where intended on a full-resolution export — storing pixels ties the annotation to whatever size the preview happened to be, which breaks the moment the window is a different width. Drags are normalised into positive-sized boxes before use, so dragging up and to the left produces a rectangle rather than a negative width that draws nothing. Stroke width scales with the image, because a two-pixel outline that reads clearly on a 800-pixel screenshot is a barely visible hairline on a 4K capture. The optional translucent fill is drawn under the stroke at about a third opacity, which highlights an area without hiding the interface underneath — the thing being pointed at is usually the thing that needs to stay readable. Annotations are drawn into the image rather than kept as a separate layer, so the exported file shows them in every viewer and cannot be toggled off. Output keeps PNG as PNG, which matters for screenshots: re-encoding screen text as JPEG produces exactly the ringing artefacts around letter edges that make a screenshot look degraded.