PDF invert rasterises, so photographs invert too

Colour invert renders each page, flips every pixel, and rebuilds from those bitmaps: text is no longer selectable, and photos, logos and chrome invert with the type because there is no object-aware pass.

PDF Colour Invert is a reading invert, not a layout editor. The tool paints the page onto a canvas, subtracts each colour channel from 255, JPEG-encodes that bitmap, and draws the JPEG back onto a new page of the original point size. Live operators do not survive. Neither do the photographs. If a product shot, a logo, or a UI screenshot has to stay as shot, crop it out before you invert, or do not invert that page.

Invert is a bitmap of the page

The file is read into memory on your machine. Each page is rendered at a scale you pick — 1.0 to 2.5, shown as 72–180 DPI. The canvas is filled #ffffff first, then the page is drawn, then every RGB sample is flipped. Alpha is left alone. The inverted canvas is encoded as JPEG at quality 0.85 and embedded. The download is named from the original plus -inverted.pdf.

That sequence is the whole product. There is no colour-operator rewrite, no “text layer only,” no object tree that knows a photograph from a heading. A raster pass cannot spare the product shot because it never saw a product shot — it saw pixels.

Guards: PDFs up to 150 MB and 500 pages. Rendering allocates about 25 MB per page at export resolution. A 500 MB scan is a laptop-memory problem. Network off, the invert still runs. This page is the pixel boundary: photographs invert because the pass is not object-aware.

Photographs invert because every pixel flips

A heading that was near-black becomes near-white. That is the reading invert people want. A photograph of a person, a pack shot, a logo on a coloured ground, a screenshot of a dashboard — those are also pixels. 255 − R, 255 − G, 255 − B. Skin goes cyan-green. A red logo goes cyan. A night photograph goes paper. There is no “text only” invert without understanding the page’s objects, which a raster pass does not.

Logos and UI chrome invert with the type. A navy mark on white becomes a pale inverse on black. If the proof still needs those objects as shot, this invert is the wrong tool for that page. The on-tool copy says it before you run. That is not a disclaimer you can prompt away.

White fill then invert has a second consequence. Transparent regions were painted paper before the draw. After the flip they are black. That is the same class of fill as JPEG conversion painting transparency white — a hole that had to become a colour — except here the colour is then inverted. You get a black rectangle where the hole was.

JPEG-backed rebuild, not live colour operators

The inverted page is a JPEG sitting on a PDF page. Glyphs are gone. Copy-out fails because the letters are not in the file. Search fails for the same reason. Inverting live PDF operators would need a full interpreter that rewrites every colour, pattern, and image. This tool does not pretend it did that.

Redaction on the same shelf is a different raster job. PDF Redaction burns boxes into a bitmap and writes PNG, because JPEG ringing at the edge of a black box can carry faint traces of the original pixels. Invert writes JPEG because the job is a reading invert, not a forensic cover. Sharper invert is a higher scale, not a PNG.

Higher scale is a sharper bitmap and a heavier file. It is not a 4K generate. Do not paste pages into an image model and ask it to “invert only the text, leave the photos.” You will spend a still rate to invent a visualisation of a proof. You will not get this tool’s pixel invert, and you will not get a live text layer.

Crop photos first, or export without invert

If the document is type plus photographs, split the job.

PDF to Images renders each page to PNG or JPG at a resolution you choose, without flipping a channel. Use it when you need the photographs as shot — a pack shot, a logo — and you will invert only the type pages, or you will not invert at all. 144 DPI is enough for on-screen review. 216 DPI or higher when small type has to stay legible.

Crop photographs out of the PDF first if they have to stay as shot, then invert the remainder here. Do not invert a catalog spread and then complain that the product shot is a negative. That is the raster doing what a raster does.

The rest of the free tools shelf is the same class of work: File API, Canvas, on your device. Invert is the one that looks like a harmless night-mode and quietly negatives every photograph on the page. Keep the inverted file for reading, not as a selectable archive.

FAQ

Why are photographs inverted too?

Every pixel is inverted, including photos, logos, and UI chrome. There is no text-only invert without understanding the page’s objects, which a raster pass does not. Crop the photos out first if they have to stay as shot.

Why can’t I select text in the result?

The page is a bitmap. The invert is a JPEG of that bitmap drawn onto a new page. Live operators are gone. If you need a selectable inverted copy, that is a different (and much harder) job, and this tool will not pretend it did it.

Is invert the same as PDF to Images?

No. PDF to Images renders pages as PNG or JPG without flipping channels. Invert renders, flips every RGB sample, and rebuilds a PDF from those JPEGs. Use to-images when the photograph has to stay as shot. Use invert when you want a dark-mode reading copy and you accept the negatives.

Should I invert, then redact the photographs I did not want flipped?

No. Invert first has already destroyed the as-shot pixels. Crop or export the photographs before invert. PDF Redaction burns boxes into a PNG raster so underlying text is gone — a different job, not a way to un-invert a pack shot.