Guide
One sentence carries this page: on a dark interface a solid fill still removes the content completely, but it stops announcing itself — so the work you lose is the reader's understanding and your own ability to check, and both have to be replaced deliberately.
Coverage and legibility are different properties
A redaction does two jobs at once, and because they almost always succeed together, most people never notice they are separable.
The first job is coverage. The pixels that held the value are replaced with something that does not encode it. This is a property of the file, it is measurable, and it either happened or it did not. The second job is legibility — not of the hidden value, but of the redaction itself. The mark has to be obvious enough that a person opening the image reads it as something was removed here rather than nothing was here. This is a property of how the picture looks to a human on whatever screen they happen to be using.
On a light interface, one black rectangle satisfies both at once. There is no realistic way to fill a region with near-black on a white panel and have a reader fail to notice. That coincidence is why redaction advice almost never separates the two ideas, and why the vocabulary for the second one barely exists.
Dark-mode captures break the coincidence. Fill a region of a near-black terminal with a near-black rectangle and coverage is total while legibility is close to zero. Depending on the particular theme, the fill may land slightly darker than the background, slightly lighter, or at a difference your eye will not resolve at all in normal viewing conditions — and that outcome varies with the panel the box happens to sit on, so two boxes in the same screenshot can behave differently. You cannot predict it from the theme name. You have to look, and looking is exactly the thing that is unreliable here.
What an invisible mark actually costs
Three distinct costs, and they land on different people.
The reader concludes the field was empty. This is the expensive one, because it produces a confident wrong belief rather than a question. A covered row in a dark table looks like a row with no value in it. A covered header looks like a panel that had not loaded yet. A reader who thinks nothing was there does not ask you what was there, so you never find out that they are working from a picture you did not send. A visible rectangle at least prompts the question.
You lose your own final glance. Most people's real verification step is a half-second look at the finished image: four boxes, four things covered, done. That glance is doing more work than anyone credits, and on a dark capture it returns nothing. You cannot count marks you cannot see, which means a region you believed you covered and a region you actually covered become indistinguishable at a glance.
Near-misses stop being obvious. A box that fell three pixels short of the end of a value leaves a fragment, and on a light capture that fragment is dark type against white with a hard black edge beside it — conspicuous. On a dark capture the same fragment is faint light type at the edge of a slightly-different-black rectangle, which is precisely the kind of low-contrast detail human vision is worst at. The error rate does not change. Your chance of catching it does.
Why your eye is a worse judge on a dark capture
There is a second, subtler problem, and it concerns blur and pixelation rather than the solid fill.
Judging whether a cover worked means judging whether you can still read what is underneath. That judgement depends on contrast, and dark interfaces often present content at much lower contrast than light ones: mid-grey text on dark grey, dim secondary labels, translucent overlays, glow effects behind type. Run a blur over a low-contrast region and it goes visually flat almost immediately — it looks finished, because there was not much visible difference to destroy in the first place. Run the same blur over black-on-white text and you get obvious grey smears that still read as words, so you keep going or switch methods.
In other words, the method that destroys least gives you the most reassuring result exactly where content is dimmest. The blur in the editor above works by downsampling the selected region by a factor of ten and scaling it back up, which means the region still contains an averaged version of what was there. Averaged signal is residual signal. Raising brightness or stretching contrast on a flat-looking dark blur can bring structure back out of it, and that is a thing a recipient can do by accident in an image viewer with auto-levels turned on.
Pixelation in this editor takes one sampled pixel per block and paints the whole block with it, so most of the data genuinely is discarded — but on a dark panel the resulting blocks can all look like the same black to you, which makes the texture that normally signals this was pixelated disappear along with everything else. You end up with a cover that is strong but, again, silent.
Solid fill does not have the strength problem at all. It is still the right method for anything you actually need gone. It only has the announcement problem, and announcements can be fixed with words.
A workflow for dark-mode captures
1. List the regions before you start. Write down, or at least count, what has to go: the account name in the title bar, the token in line four, the two e-mail addresses in the sidebar. Four items. You are replacing a visual check with an arithmetic one, so you need the number to exist before you begin, not after.
2. Use Black Box, not blur, for anything sensitive. The safety banner that appears when you pick Blur or Pixelate is not a formality, and the contrast argument above makes it more important on dark captures rather than less.
3. Overshoot the edges on purpose. You will not spot a three-pixel miss on this image, so make the margin larger than the one you would use on a light screenshot. Extending a box into surrounding chrome costs nothing. On dark captures an overshoot has a second benefit: a rectangle that crosses a boundary between two differently-dark surfaces reveals itself geometrically, because it cuts a straight edge across a border that used to continue.
4. Verify against the exported file, not the canvas. Download the PNG, open it, zoom well past one hundred per cent, and raise the brightness or stretch the contrast before you decide a region is clean. The full procedure is on how to verify a photo was actually redacted, and every step in it matters more here than on a light capture, because none of the shortcuts work. Note also that the editor scales large images down to fit the page, so what you are looking at while you draw is often smaller than the file you are about to send.
5. Count, do not look. Four regions on the list, four covers drawn. If the count does not match, undo back and redo rather than squinting. Undo steps back one cover at a time; Clear is not a comparison toggle, it throws away every cover you have made and empties the undo history with them.
6. Say what you covered. Because the mark does not speak for itself, a sentence has to. Name the number of regions and the class of information in each — four covered areas, two of them e-mail addresses in the sidebar, one an access token, one an account name — and the reader can no longer mistake a removal for an absence. Keep the description at the level of the field, never the value.
When to re-capture in light mode instead
Sometimes the better fix is upstream. If the image is going somewhere the distinction between removed and empty genuinely matters — a dispute, a formal complaint, anything a counterparty will read, anything published where you cannot answer follow-up questions — switching the application to its light theme and taking the screenshot again costs one minute and removes the whole problem. The same rectangle on a light panel does its own explaining.
This is worth it when the reader has reason to wonder whether you removed something inconvenient, when the image will be forwarded on past the person you are sending it to, when it will be printed (a dark screenshot with near-black covers is the worst possible thing to put through a printer, and the marks are the first casualty), and when the picture is one of a set that somebody will have to read quickly.
It is not worth it when the view cannot be reproduced — a transient error, a notification that has gone, a state you would have to break something to get back — or when the capture is going to one colleague who already knows the context. In those cases keep the dark image, use solid fills, and carry the explanation in the message instead.
Which method, given the constraint
A short way to decide, given that this editor's fill colour is not adjustable. If the content is sensitive, use Black Box — always, dark theme or not — and solve visibility with a sentence or a re-capture. If the content is not sensitive and you only want to de-emphasise it, the trade-offs between the three methods are set out in blur vs pixelate vs black box, with the dark-capture caveat that your own impression of how much a blur destroyed will be too generous. And if what you need is for the reader to see clearly that a specific region was deliberately removed, no setting in any region-based editor achieves that on a near-black panel. That job belongs to the text next to the image.
What the editor above does and does not do
The tool on this page is region-based: three methods — Black Box, Blur, Pixelate — across three selection shapes, rectangle, oval and freehand lasso. Each operation replaces the pixels inside the region you drew. Black Box fills it with a fixed #111111; Blur downsamples the region heavily and scales it back up; Pixelate replaces each block with a single sampled colour. Undo steps back one cover, Clear restores the untouched image and discards all of them, and Download PNG writes the current canvas out as hideshot-<timestamp>.png, re-encoded from scratch rather than copied from the source file. Everything happens in your browser on your own machine; nothing is sent anywhere, which is also why there is no stored copy and no sharing link.
Three limits matter specifically for dark captures. The fill colour is fixed — there is no colour picker, no white or red box, no outline or hatch option, so you cannot make the mark contrast with its background from inside this editor. There is no text, arrow, label or caption tool, so the explanation cannot be written onto the image here; it belongs in the message or the ticket comment. And there is no invert, brightness or levels control, so the contrast check that this page keeps recommending has to be done in an image viewer or editor after you export, not on the canvas above.
Common mistakes and misconceptions
"If the box is invisible, the text might still be visible." No. These are unrelated. A solid fill replaces every pixel in the region with the same value, and a region of uniform colour carries no information about what it replaced regardless of how it compares with the surrounding interface. The hidden value is gone. What is missing is the signal that it was ever there.
"I will use blur instead, since a black box blends in anyway." This is the worst available move: it trades the strongest method for the weakest in order to solve a communication problem, in the exact conditions where low contrast makes a weak blur look convincing. Keep the solid fill and add a sentence.
"The dark screenshot is hard to read anyway, so it is safer." Hard for you, on your screen, at the brightness you happen to be using. Brightness, gamma and contrast are all adjustable at the other end, and plenty of viewers apply automatic levels without being asked. Difficulty of reading is not a security property.
"A reader will obviously understand that a dark rectangle is a redaction." They would need to notice it first. Panels that look empty in dark themes are entirely normal, which is why an empty-looking region is the interpretation that costs the reader nothing to believe.
"I will check it after export by eye, like always." Checking by eye is the step that stopped working. Zoom in and raise the brightness, or check the count against a written list — and do it on the file you are about to send rather than on the live canvas.
"Inverting the image would expose what is under the box." It would not. Inversion, brightness and levels are tone mappings; they cannot recover information from a region where every pixel is identical. They can expose structure inside a blurred region, which is a reason to prefer the solid fill, not a reason to distrust it.
"Switching to light mode and re-shooting is overkill." It is a minute of work that replaces every workaround on this page, and for anything a counterparty will read or anybody will print, it is the cheapest fix available.