Free · No Signup

Redacting Images for a Public Records Request

Every other page here helps you disclose as little as possible — a records law obliges you to disclose as much as possible, mark how much you took out, and name a reason for each box.

🔒 No upload · Runs in your browser · Instant download

Every other page on this site is written for someone trying to give away as little as possible. You hold an image, you decide what the recipient does not need, you cover it, and the only judge of whether you went too far is your own sense of what the reader has to do with the file. Covering more is always the cautious choice.

Answering a public records request inverts almost all of that. You are on the release side of a statute, which means the default is disclosure and every black box is an exception you have to justify. The duty is to release everything reasonably segregable, not everything you feel comfortable releasing. The redaction is supposed to be conspicuous rather than tidy, because the law asks you to indicate how much was taken out. Each box needs a citable reason attached to it, recorded somewhere a reviewer can read months later. And covering more is not the safe error — it is withholding without a basis, and it is the thing that gets appealed.

That makes this a different job from the rest of the site rather than a new document type to cover. What follows is what the federal statute actually says, how its three duties translate into a workflow for a photograph or a screenshot, the per-region log that carries the part this editor cannot put on the image, and the places where image productions behave worse than document productions. None of it is legal advice: your statute, your exemption list and your signatory decide the outcome.

Mode
Shape

Drop your image here

Or click to browse · Paste with Ctrl+V also works

PNG · JPG · WebP · GIF
How It Works
1

Open

Drop your image in or paste from clipboard.

2

Pick Mode

Black Box, Blur, or Pixelate.

3

Select Areas

Rectangle, oval, or freehand lasso — then hide what you selected.

4

Download

Hit Download PNG. Done.

SquooshNeed to shrink your image after editing? Squoosh is a free browser-based image compressor with no upload required.

Visit Squoosh →
Guide

One sentence carries this page: on a records release the redaction is not just a cover but a representation — it has to leave behind everything segregable, show how much was removed, and carry a citable reason, and this editor can only do the first two of those three.

What a records law asks of a redaction

The federal Freedom of Information Act is the clearest written version of the obligation, so it is worth quoting rather than paraphrasing. Three passages matter.

On when you may withhold at all, 5 U.S.C. § 552(a)(8)(A) directs that an agency shall withhold information only if — (I) the agency reasonably foresees that disclosure would harm an interest protected by an exemption described in subsection (b); or (II) disclosure is prohibited by law, and shall consider whether partial disclosure of information is possible whenever the agency determines that a full disclosure of a requested record is not possible and take reasonable steps necessary to segregate and release nonexempt information.

On what is left over, the closing lines of § 552(b) provide that Any reasonably segregable portion of a record shall be provided to any person requesting such record after deletion of the portions which are exempt under this subsection.

And on marking, the same passage continues: The amount of information deleted, and the exemption under which the deletion is made, shall be indicated on the released portion of the record, unless including that indication would harm an interest protected by the exemption in this subsection under which the deletion is made. If technically feasible, the amount of the information deleted, and the exemption under which the deletion is made, shall be indicated at the place in the record where such deletion is made.

Read those together and a records redaction has three duties rather than one: segregate and release, indicate the extent, and name the exemption. Nothing on the rest of this site asks for the second or the third.

One scope warning, and it is a real one. That is the federal statute, binding on federal agencies. If you are a county clerk, a police department, a school district, a water authority or a contractor answering a state request, your own public records law governs: the exemption list is different, the marking requirement may be stronger, weaker or absent, and so is the appeal route. Take the passages above as the shape of the obligation and check your own statute, your agency's manual or your records counsel for the content. This page is about what an image editor can and cannot do for you; it is not legal advice, and the person who signs the response decides.

Duty one: segregate, do not bulk-withhold

For a photograph, segregability is spatial. The exempt material occupies some part of the frame — a bystander's face, a complainant's house number, a case file number on a monitor — and the rest of the frame is the reasonably segregable portion you owe the requester. The statutory instinct therefore runs opposite to the instinct this site usually trains: cover the face and release the photograph, rather than withholding the photograph because there is a face in it.

Two edges to that. If the exempt thing genuinely is the whole frame — the entire image is a person's medical chart, or the subject of the photograph is itself the protected fact — then there is no segregable portion, and the honest response is a withholding with a stated basis, not a wholly black rectangle produced as though it were a record. And if the exempt material is distributed across the picture rather than sitting in a place, a box cannot reach it; the identifier that is not shaped like a region is a recurring problem with its own treatment, and the answer there is usually to withhold the image rather than publish a cover that does not work.

Duty two: let the box show its own extent

This is the one place on this site where you want the redaction to be loud. The visible mark is your indication of the amount deleted, so it has to read unmistakably as a deliberate deletion.

That rules out Blur and Pixelate on a release copy, and not only for the usual reason. Blur in this editor downsamples the selected area by a fixed factor of ten and scales it back up; Pixelate fills each block with one sampled pixel, using a block size of the selection's shorter side divided by twelve with a six-pixel floor. Both preserve coarse structure by design, so a face can stay recognisable as a shape. But the compliance problem is separate: a softened region does not tell the reader that anything was removed. It looks like a bad photograph. A flat #111111 fill with a hard edge announces itself, and that self-announcement is the part the statute is asking for.

The corollary is that box size is a representation too. An oversized rectangle swept across half the frame overstates what was exempt and invites the question of what else went under it; a box drawn tight inside the edge of a face understates the deletion and risks leaving a jaw line or a hairline outside it. Aim for the exempt material plus an honest margin, consistently, across every image in the production.

One mechanical trap in this editor matters more here than anywhere else: a rectangle drag under about four image pixels, or an oval under six, is discarded silently. A small box over a case number or a date of birth can simply fail to appear, with no warning and no visible difference until someone reads the produced file. Confirm every box is actually on the canvas before you download, and remember that the canvas is displayed scaled to fit while the edits land at the source's full pixel dimensions.

Duty three: the exemption cannot go on the image

Here is the hard limit, stated plainly. This editor cannot write the exemption at the place of deletion. There are three modes and three shapes; every operation replaces the pixels inside a path you draw; there is no text tool, no annotation layer, no caption and no colour choice, and the page's editor code contains no text-drawing call whatsoever. No amount of technique will put (b)(6) beside a box.

Federal FOIA conditions the place-of-deletion marking on technical feasibility, and feasibility is a property of your toolchain rather than a reason to drop the requirement. So the citation has to travel next to the image. Three workable arrangements, in increasing order of effort:

A redaction log plus the response letter. The letter names the exemptions relied on; the log gives one row per covered region so the reader can tie each box to a basis. This is the minimum, and for most image productions it is enough.

The image placed in a document with captions. Drop the redacted PNG into a word processor or PDF with a numbered caption under each figure and the exemption code in the caption. The deletion is still done here, destructively, at the pixel level; the labelling is done where text is possible.

A stamping step. If your practice or your statute requires the code adjacent to each mark on the face of the record, you need a tool that writes text onto the page after the pixels are gone. Do the destruction here first, so the stamping tool never holds the unredacted pixels in a layer somebody can peel off.

Describing what you removed in writing has a failure mode of its own — a note can disclose exactly what the box was hiding — and that trade-off is worked through in explaining what you redacted without undoing the redaction. For a records response the usual answer is to name the category and the exemption, never the value.

A redaction log for images

Document productions get a log almost for free, because pages are numbered and text can be quoted. Images need one built by hand. One row per covered region, not per file:

Produced filename. This editor exports as hideshot-<timestamp>.png, so rename it to your production naming scheme before it leaves your working folder — and do not encode the withheld content in the new name.

Source file and where the unredacted original sits. The production copy is not the record; the original is. Retention and preservation of records under request is the records officer's call and your statute's, not a step in an image editor.

Region number and rough location in the frame. “Three faces along the counter, upper left”, “account number, lower right panel”. Because an image has no words or lines, the natural unit for “amount deleted” is a count of regions plus a description of each.

Class of information covered. The category, never the value. “Third-party name”, not the name.

Exemption cited. Per region. A single incident photo can easily carry a bystander's face, a complainant's address and an internal case number under three different bases.

Mode used, decided by, date. Black Box throughout, if you follow this page. The initials and the date are what let someone defend the response a year later.

The log earns its keep in two moments. First, on appeal or review, when you will be asked what was under each box and need to answer without producing the original. Second, when a different requester asks for the same records later — because if you redraw the boxes freehand on a second production, the two releases will not agree, and the pair discloses the union of what either one left visible. Consistency across productions is a logging problem, not a drawing problem.

Why image productions behave worse than document productions

Screenshots of internal systems carry chrome. The record requested is the case note; the capture also contains the operator's username, a workstation name, an internal URL with a record identifier, a clock that fixes the time zone, and a row of notification badges. Little of it is responsive to the request and some of it may be exempt on security grounds, but blanket-covering the whole periphery can look like you hid substance. Decide deliberately, and log the decision.

The frame is preserved, which is useful. There is no crop, resize or rotate anywhere in this editor, and the canvas is created at the source image's exact pixel dimensions, so the file you download is the same frame at the same size as the source with painted regions inside it. That is an easy property to state in a response and an easy one for a reviewer to verify.

Metadata does not survive, which cuts both ways. Download PNG serialises the canvas bitmap, so the produced file carries none of the source's EXIF or provenance fields. Good news if the camera recorded GPS coordinates you had not planned to release; worth a note if your records system expects a produced copy to retain capture metadata. Either way, the original keeps its metadata and stays in the file.

Faces are the common case and the hardest to bound. Agencies releasing stills from body-worn or fixed cameras run into uninvolved bystanders constantly, and that specific situation — which faces are incidental, how much of the surrounding frame identifies them anyway — is covered in more detail in blurring faces in body cam footage. On a records release, use Black Box there too, for the extent-marking reason above.

Clear has no confirmation. The Clear button repaints the retained original over your work immediately, with no prompt, and it wipes the undo history with it. Forty boxes into a long screenshot that is an expensive misclick, and there is no autosave anywhere, because nothing is persisted outside the tab.

Common mistakes and misconceptions

“There is a face in it, so withhold the photo.” Segregability points the other way: release the frame, cover the face, and record why.

“Over-redacting is the safe error.” Not on this side of the transaction. Withholding beyond the exemption is withholding without a foreseeable-harm basis, and it is the most common thing an appeal is about.

“Blur is gentler and still counts.” Gentler is the wrong axis. A factor-of-ten downsample can leave a recognisable shape, and a soft patch does not indicate to the reader that anything was deleted at all.

“The box is the record of what I did.” The box shows extent. It does not show the reason, the category or who made the call, and without those the response cannot be defended by anyone except the person who drew it.

“I will redact it and delete the original so it cannot leak.” Destroying a record that is under request is the one move in this whole workflow with serious consequences. Custody of the original is a records decision, and the correct answer is almost always to keep it where it already lives.

“Marking the exemption on the image is optional.” Federal FOIA asks for it at the place of deletion where technically feasible, and on the released portion of the record regardless, subject to the harm proviso. This tool cannot do the on-image part, so another step in your chain has to.

“A second request for the same photo can just be redacted again.” Redraw by hand and the second set of boxes will not match the first. Two productions of one photograph release everything that either version left visible.

“The requester will not zoom in.” Assume the opposite. A released image is a published image: it will be inspected at full resolution, enhanced, and compared against the other copies already in circulation.

On a Records Release, the Redaction Has to Justify Itself

Redacting an image for a public records request is the one job on this site where covering more is not the cautious option. Everywhere else you are trying to disclose as little as possible, and the only test is whether the recipient can still do what they need to do. On the release side of a records statute the default is disclosure, and each black box is an exception you have to be able to defend. The federal Freedom of Information Act puts the obligation in three parts. Under 5 U.S.C. § 552(a)(8)(A) an agency shall withhold only if it reasonably foresees that disclosure would harm an interest protected by an exemption, or disclosure is prohibited by law, and shall consider partial disclosure and take reasonable steps to segregate and release nonexempt information. The closing lines of § 552(b) add that any reasonably segregable portion of a record shall be provided after deletion of the exempt portions, that the amount of information deleted and the exemption relied on shall be indicated on the released portion of the record unless that indication would itself cause the protected harm, and that the same shall be indicated at the place in the record where the deletion is made if technically feasible. Those quotations are the federal statute; a city, county, school district or state body answers to its own law, with its own exemption list and its own marking rules.

For a photograph or a screenshot, segregation is spatial: the exempt material sits in part of the frame and the remainder is what the requester is owed, so the statutory move is to cover the face and release the picture rather than withhold the picture because of the face. The extent requirement then changes which mode you should use. Blur in this editor is a fixed factor-of-ten downsample scaled back up, and Pixelate uses blocks of the selection's shorter side divided by twelve with a six-pixel floor, so both can leave a recognisable shape — but the decisive problem is that a softened patch does not read as a deliberate deletion. It reads as a bad photograph. A flat #111111 Black Box with a hard edge announces its own boundary, which is exactly the indication the statute asks for. Box size is part of that representation too, so keep each one to the exempt material plus an honest margin, and check every box actually landed: a rectangle drag under about four image pixels, or an oval under six, is discarded silently by this editor.

The third duty is the one this tool cannot help with, and that is worth saying before you plan a production around it. There is no text tool, no annotation layer and no caption field here — the editor code contains no text-drawing call at all — so the exemption code cannot go on the image. It travels beside it instead: a redaction log with one row per covered region, the exemptions named in the response letter, and, where your practice requires a code adjacent to each mark, a stamping step after the pixels are already gone.

The log is what turns a set of black boxes into a defensible response. One row per region, not per file: produced filename (this editor exports hideshot-<timestamp>.png, so rename it into your production scheme), the source file and where the unredacted original stays, the region's rough position in the frame, the class of information covered rather than its value, the exemption cited for that region, and the mode, the initials and the date. It pays for itself twice — on appeal, when someone asks what was under each box and you answer without producing the original, and on the next request for the same records, because freehand boxes redrawn months later will not match and the two productions together release whatever either one left visible. A few image-specific quirks are worth logging around: a screenshot of an internal system carries usernames, workstation names, internal identifiers and a clock along with the record itself; the export is the source's exact pixel dimensions because there is no crop, resize or rotate anywhere in the editor; the PNG is written from the canvas bitmap, so no EXIF or provenance travels with the produced copy; and Clear repaints the original with no confirmation and no way back. Nothing is uploaded and nothing is stored — the editor has no fetch call, no XMLHttpRequest, no WebSocket and no browser storage, so the records you are processing never leave the machine you are sitting at. None of this is legal advice: your statute and your signatory decide the response.

Frequently asked questions

Is covering a face in a photo enough to meet a segregability duty?

Covering the face is the point of the duty rather than a shortcut around it. Under the federal Freedom of Information Act, 5 U.S.C. 552(b) provides that "Any reasonably segregable portion of a record shall be provided to any person requesting such record after deletion of the portions which are exempt under this subsection", and 552(a)(8)(A) directs an agency to consider whether partial disclosure is possible whenever full disclosure is not, and to take reasonable steps necessary to segregate and release nonexempt information. For a photograph, segregation is spatial: the exempt thing occupies part of the frame, and the rest of the frame is what the requester is owed. So covering the face and releasing the photo is the shape the statute asks for, and withholding the whole image because a face appears in it is the move that draws an appeal. What covering the face does not do is discharge the other two duties set out in the same part of the statute - indicating the amount of information deleted, and the exemption under which each deletion was made. Those are separate obligations, and a black box states neither a reason nor a category. Note too that the quoted text is the federal statute; if you work for a city, county, school district or state body, your own public records law has its own exemption list and its own marking and appeal rules, and the person who signs the response makes the call.

Can I write the exemption code next to each black box in HideShot?

No, and it is better to know that before you plan a production around this tool. The editor on this page has three modes and three shapes and does exactly one kind of thing: it replaces the pixels inside a path you draw. There is no text tool, no annotation layer, no caption field and no colour control - the editor code contains no text-drawing call at all - so there is no way to put a code such as (b)(6) onto the image itself. Federal FOIA asks that the amount of information deleted and the exemption be indicated at the place in the record where the deletion is made "if technically feasible", and technical feasibility is a property of your toolchain rather than a licence to skip the marking. In practice the citation travels beside the image instead: a redaction log with one row per covered region, the exemptions named in the response letter, and - where your own practice requires the code on the face of the record - a second step in something that can stamp text, with the image placed into a document or PDF. HideShot's job in that chain is destroying the pixels, which it does properly and irreversibly.

Should I blur faces instead of blacking them out on a release copy?

No, for two reasons, and the second is the one that gets missed. First, Blur in this editor is a fixed factor-of-ten downsample of your selection scaled back up, and Pixelate replaces each block with a single sampled pixel, its block size being the shorter side of the selection divided by twelve with a floor of six pixels. Both are designed to discard fine detail while preserving coarse structure, so a recognisable head shape, hairline and posture can survive either one. Second, and specific to a records release: a blurred face does not announce that anything was removed. A reader cannot distinguish a deliberate deletion from a soft photograph, and the statute asks you to indicate the amount of information deleted on the released portion of the record. A flat #111111 rectangle with a hard edge is self-announcing - the extent of each deletion is visible at a glance, which is precisely what you want on this one kind of image and almost nowhere else on this site. Use Black Box for everything you withhold, and keep each box to the exempt material plus an honest margin, since the size of the box is itself part of what you are representing about the deletion.

Does HideShot send the records I am redacting anywhere?

No, and for records work you can verify that rather than take it on trust. The editor code on this page contains no fetch call, no XMLHttpRequest and no WebSocket, and no address of any kind, so there is nothing in it capable of transmitting your image. The path a file takes is short and entirely local: the file picker, the drop zone or Ctrl+V hands it to your browser, the browser decodes it into an image, that image is drawn onto a canvas created at its exact pixel dimensions, and every Black Box, Blur or Pixelate operation is arithmetic on those pixels. Download PNG asks the canvas for its contents and hands those bytes to the browser as a data URL, so the saved file comes out of your own tab. Nothing is persisted either - no account, no saved settings and no browser storage of any kind - so the image, the undo snapshots and the retained original live in that tab's memory and go when it closes. Two consequences for a records officer: the file never leaves the machine you are sitting at, which matters when the records cannot leave your network; and if the tab closes mid-production, you start that image again.