A wrong-looking result from a PDF-to-grayscale conversion usually isn't a bug — it's the tool doing what it says: building a new raster PDF from your pages using the formula round(0.2126 × R + 0.7152 × G + 0.0722 × B) per pixel.
The Grayscale PDF Converter reads your local PDF in the browser, renders each bounded page on a canvas, replaces the red, green, and blue channels of every visible pixel with a single computed gray value, encodes that canvas as a PNG, and stitches the PNGs into a fresh PDF. Because the output is rebuilt from rasterized pixels rather than recoloring the original PDF's operators, what you see in the new file is the page's visual appearance only — not its text layer, link annotations, form fields, vector paths, or accessibility structure. Many "wrong" results turn out to be the expected outcome of that pipeline: thicker-looking type, smoothed edges, larger file size, or missing searchability. The fix is rarely to convert again with a different button — it's to confirm the source, confirm the page limits, and re-run with realistic expectations about what the tool preserves.
Why the Grayscale PDF Converter Output Looks Different
The Grayscale PDF Converter doesn't recolor the operators inside your PDF. The browser-side loading and rendering is handled through a JavaScript PDF library that draws each page onto an HTML canvas (see the Mozilla PDF.js API reference for the underlying renderer). Once a page is drawn, the tool walks every pixel of that render, and a fresh PDF is built from PNG page images. The visible content is faithful to the source, but anything that lived in the original PDF's structure — text, links, form fields, vector strokes, layers, accessibility tags, signatures, and the original compression — is dropped, because those structures were never on the canvas in the first place. Two consequences explain most "wrong-looking" results:
- Fine type and thin strokes look softer than the original. The page is rasterized at a fixed 1.5 scale and placed back onto a page of the original size, so edges that were crisp in vector form now depend on how the browser's renderer sampled them. Body type below about 8 pt often picks up visible softness.
- The grayscale tones don't match what you'd get from a printer's color profile or from a "Convert to Grayscale" option in a desktop editor. The tool uses one fixed luminance formula on every pixel, so a saturated red and a saturated blue that look similarly bright to your eye can land on noticeably different gray values. That's not a bug — it's the formula's weighting.
If your comparison point is "the same look as the source, just desaturated," the result will almost always feel off. If your comparison point is "a clean monochrome copy of the page's appearance, with text no longer selectable," the result is doing what it's designed to do. Reviewing a related guide on checking the result after a grayscale conversion can also clarify which "problems" are really expected tradeoffs.
The Per-Pixel Formula Behind Every Grayscale Value
The Grayscale PDF Converter applies the same arithmetic to every visible pixel: it multiplies red by 0.2126, green by 0.7152, and blue by 0.0722, sums the products, rounds to the nearest 8-bit value, and writes that single number into all three RGB channels. That's the Rec. 709 luma weighting, and it's the same family of formula browsers expose through the canvas filter property — but this tool does the math in code rather than relying on that filter, because the filter is not a dependable default in Safari. The per-pixel reading is done through the standard canvas pixel API (see MDN's CanvasRenderingContext2D getImageData reference), so the behavior is testable rather than dependent on browser-specific filter quirks.
The weighting explains a few things you might notice in the output:
- Greens get more visual weight than reds or blues. A leaf-green pixel that looks medium-bright in color becomes a fairly bright gray.
- Blues darken the most. A pure blue pixel (R=0, G=0, B=255) lands at 0.0722 × 255 = 18.4, rounded to 18 — nearly black.
- Already-gray pixels are unchanged. A mid-gray pixel (R=128, G=128, B=128) produces 128 from any of the three channels and stays at 128.
- Pure red darkens more than pure green. A pure red pixel produces 0.2126 × 255 = 54.213, rounded to 54 — visibly darker than a green that lands in the 180s.
A single worked example, shown step by step so you can verify it: a pure red pixel, R=255, G=0, B=0. The formula gives 0.2126 × 255 + 0.7152 × 0 + 0.0722 × 0 = 54.213, rounded to 54. The output pixel is RGB(54, 54, 54). That's why reds in the converted document often look darker than you expect — they really are dark grays under this weighting.
How to Fix a Wrong-Looking Grayscale PDF
Work through these checks in order before you convert again. Most "wrong" outputs are fixed by understanding what the tool is supposed to produce, not by retrying with different clicks.
- Open the source PDF and the grayscale output side by side. Look at the visual appearance of the page, not at whether text is selectable or links work. If the monochrome image matches the source's layout, the conversion did its job.
- Confirm the file is under 25 MB and contains 40 pages or fewer. Files outside those limits are rejected before any page is rendered, so a missing output usually means the preflight check stopped the job rather than a bug in the conversion.
- Confirm none of the per-page budgets are exceeded at the 1.5 render scale. The tool checks each page for 12,000 pixels on the longer side, 40 megapixels per page, and 100 megapixels across the whole document. A single oversized page or a wide two-page spread can hit these caps and cancel the run before any output is written.
- Check whether the source is password-protected, digitally signed, malformed, or unusually complex. The tool does not bypass passwords, validate signatures, or claim archival conformance, and any of these conditions can stop the load step. Remove the password first or use a different tool that handles signed files.
- Decide whether you actually need the source's interactive features in the output. If selectable text, form fields, live links, or accessibility tags matter, the raster result will never satisfy you — the fix is a different tool, not a re-run.
- If everything checks out and the visual still looks wrong, close the tab and start a fresh conversion. The previous job's canvases, PDF worker tasks, and temporary download URLs are released on cancellation, and starting over with a clean state avoids carrying stale buffers between runs.
For a structured list of pitfalls to watch for, the guide on common mistakes when using a PDF-to-grayscale converter lines up with most of the items above and is a useful companion while you debug.
Limits and Budgets That Stop a Conversion Before It Starts
The Grayscale PDF Converter preflights the file before it allocates a single output canvas. If any of the checks fail, the job is rejected up front and the source is not modified. Knowing these numbers tells you whether a "wrong" output is really a "rejected" output.
| Limit | Value | What it controls |
|---|---|---|
| Source file size | 25 MB | Largest accepted input file |
| Page count | 40 pages | Largest accepted document length |
| Rendered page side | 12,000 px | Longest edge of one page at the 1.5 render scale |
| Per-page pixels | 40 megapixels | Raster size of one page |
| Document pixels | 100 megapixels | Total raster size across all pages |
| Render scale | 1.5× | Fixed scale applied before page is placed back at original size |
These limits keep one browser tab from trying to hold an unbounded image buffer. If your file is rejected, split it with the Split PDF tool first, or trim it with Delete PDF Pages, and run each part through the Grayscale PDF Converter separately.
When the Result Is Correct and Just Unfamiliar
A few "problems" are baked into the design:
- Selectable text disappears. The output is a PNG-on-PDF page, so the new file has no text layer at all. Searching inside the result will return nothing — search inside the source if you need that.
- Links, form fields, annotations, layers, and signatures don't carry over. They are not part of the raster canvas, and there is no structural layer in the new file for them to live on.
- The file gets larger. PNG page images can use more bytes than the original PDF's fonts, vectors, and image compression. A 2 MB color PDF can easily become a 15 MB or 20 MB grayscale raster PDF.
- Page dimensions match the source but the visual detail is rendered at a fixed 1.5 scale. That trade keeps the tool fast and bounded, but it means a page that contained 4-pt vector type in the original will not be as crisp after conversion.
If the result is a clean monochrome copy of the page's appearance and the limits above were respected, the output is correct. The mismatch is in expectations, not in the file. Keep both the source and the grayscale output, and inspect the grayscale file page by page before you share or print it.
When to Reach for a Different Tool Instead
Stay with the Grayscale PDF Converter when you want a quick visual monochrome copy, a print preview, a file that meets an upload requirement that demands grayscale, or a reference document where flattened pages are acceptable. Switch tools when you need any of the following:
- Selectable text and live links in the output. Those don't survive a raster rebuild, and no browser-side re-run will bring them back.
- PDF/A compliance, accessibility tags, or controlled compression. The output is not archival and is not optimized for size.
- Vector graphics preserved as vectors, or signatures validated. The raster pipeline drops vector information, and the tool explicitly does not claim signature validation or legal conformance.
- Print color management. The fixed luminance weighting is not a substitute for a calibrated print workflow, and the file is not tagged with any color profile.
If your source has interactive content that has to survive, run it through a professional PDF editor and ask that editor to apply its own grayscale conversion. The browser tool is for a different job, and the right fix for "I need my text back" is not another re-run — it is a different converter that preserves the PDF's structure.