E-Ink Image Optimization: Why Your Screensaver Looks Bad

Why images look worse on e-ink than they did on your screen

Nearly every guide to e-reader screensavers stops at the pixel dimensions. Getting those right matters — it is the difference between a sharp image and a rescaled one — but it is not why a correctly sized photo can still arrive looking muddy, striped or flat.

An e-ink panel is a fundamentally different display from the one you edited the image on, in three ways that each change what you should send it.

How many shades of grey does e-ink actually show?

Sixteen. Standard E Ink Carta panels are 4-bit devices: every pixel settles into one of 16 grey states, roughly evenly spaced from black to the panel's white. Your image has 256 levels per channel. Somewhere between the file and the screen, 240 of them are thrown away.

For most of a photograph this goes unnoticed, because detailed areas have enough local variation to hide the steps. Where it becomes obvious is anywhere the tone changes slowly and smoothly across a large area: a clear sky, a studio backdrop, a soft vignette, a gradient behind text. Those regions have nowhere to hide the quantization, so the 16 steps appear as visible bands.

This is the single most common reason a screensaver looks wrong despite being exactly the right size, and it is almost never mentioned. If you have ever put a landscape photo on an e-reader and found the sky had turned into four or five flat stripes, that is what happened.

Fixing banding: dither it yourself

The reduction to 16 greys is going to happen either way. The only question is whether you control it or the device does.

Left alone, the device does the simplest possible thing — it rounds each pixel to the nearest available level, which is exactly what produces hard bands. Dithering instead distributes the rounding error into neighbouring pixels, so a region that falls between two available greys is rendered as a fine mix of both. The banding is traded for noise, and on a 300 ppi panel that noise is invisible at reading distance.

In GIMP: Image > Mode > Indexed, generate an optimum palette with a maximum of 16 colours, and set colour dithering to Floyd-Steinberg. In Photoshop: Image > Mode > Grayscale, then Image > Mode > Indexed Color with 16 levels and Diffusion dither.

With ImageMagick, there is a subtlety worth knowing. The obvious command asks for an optimal 16-colour palette, which picks 16 greys that suit your particular image — but the panel does not have those greys. It has its own roughly evenly spaced set. Remapping to an even ramp matches the hardware better:

# build a 16-step grey ramp once
magick -size 1x16 gradient:white-black -colorspace Gray ramp.png

# then dither any image onto it
magick photo.png -colorspace Gray \
  -dither FloydSteinberg -remap ramp.png out.png

One caveat on where this helps. On a Kobo, a Boox or any device that reads your file directly, the file you copy across is the file the panel displays, so your dithering survives intact. On a Kindle, the image passes through an EPUB conversion on its way to the device, which can re-process it — so dither there by all means, but expect less control over the result.

Dithering is not universally good. Images built from flat areas and hard edges — logos, typography, high-contrast graphic art — have no gradients to band, and adding dither noise to them only makes clean areas look grainy. Reach for it when the image has smooth tone in it, not by default.

Setting contrast for a reflective panel

An e-ink screen does not emit light. Its white is room light bouncing off a surface that is, in absolute terms, light grey; its black is dark grey. The resulting contrast ratio sits somewhere around 15:1 to 20:1 depending on panel generation, against 1000:1 or more on the phone you are reading this on.

The practical effect is that the tonal range gets compressed from both ends. Deep shadows converge on a single near-black; bright highlights converge on paper-white. Whatever separation your image relied on in those regions is gone.

So push the source harder than feels right on a monitor. Set the black point so something in the image is genuinely black and the white point so something is genuinely white, then increase mid-tone contrast on top. An image that looks slightly overdone on your screen usually looks correct on e-ink — and one that looks perfect on your screen usually looks flat there.

Under a frontlight the panel gets brighter but not more contrasty, so this holds regardless of lighting. On the reMarkable 2, which has no frontlight at all, it matters more still: the screen is only ever as bright as the room it is in.

Colour is a brightness problem, not a colour problem

On a greyscale panel, colour does not become a duller version of itself — it becomes its luminance and nothing else. A red subject against a green background is unmissable in colour and can be completely invisible in grey, because those two colours can share almost the same brightness.

The test takes five seconds: desaturate the image and look at it. Anything you can no longer distinguish, the e-reader will not distinguish either. If the subject vanishes, you need to fix it in the source by changing luminance — dodging and burning, adjusting the individual colour channels' brightness before converting — not by adjusting saturation, which does nothing at all here.

Colour e-readers: Kaleido 3 and Gallery 3

Colour models change this advice, and the two panel technologies in circulation change it differently.

Kaleido 3 — the Kindle Colorsoft, Kobo Libra Colour and Kobo Clara Colour — lays a colour filter over the same greyscale grid. Greyscale still renders at the panel's full 300 ppi, but colour resolves at roughly half that. Large flat areas of saturated colour reproduce well; fine coloured detail and small coloured type go soft. Colours also read noticeably more muted than on a backlit screen, so start from source images that are already high in contrast and saturation rather than subtle ones.

Gallery 3 — used in the reMarkable Paper Pro — keeps full resolution in colour instead of halving it, so coloured detail holds up far better. Its trade-off is refresh speed, which is irrelevant for an image that sits still. It is the only device in this range where fine coloured linework is worth attempting.

On neither type does the required image size change. A Colorsoft takes exactly the same 1264 × 1680 file as the greyscale Paperwhite beside it.

Sharpening, and how much

E-ink scatters light from a diffuse surface rather than emitting it through a defined pixel aperture, and there is no sub-pixel rendering to lean on. The result is a small but real loss of apparent edge definition, which a light unsharp mask compensates for.

Keep it restrained. Sharpening halos — the bright fringe along a dark edge — are far more visible against flat e-ink greys than they are on a busy backlit screen, and once you have quantized to 16 levels a halo can land on its own distinct grey step and read as a deliberate outline. Sharpen last, after any dithering decisions, and at a lower amount than you would for print.

PNG or JPG?

PNG. JPEG is lossy, and the artefacts it leaves — ringing and mosquito noise around high-contrast edges — are largely invisible on a backlit screen but survive the reduction to 16 grey levels perfectly well. Text edges and hard graphic boundaries are where you will see them.

The usual reason to accept JPEG is file size, and that reason does not apply here: a screensaver is one image sitting on a device with gigabytes free. There is nothing to save. If you want a smaller file anyway, a PNG already quantized to 16 greys compresses extremely well — often smaller than the JPEG would have been, and with no artefacts at all.

The recipe, in order

1. Crop to the exact panel resolution. Crop and export at your device's native pixel dimensions — 1072 x 1448 for a 6-inch Kindle or Kobo Clara, 1264 x 1680 for a 7-inch model. Any rescaling on the device resamples the image and softens it before the panel ever sees it.
2. Convert to greyscale and look at it. On a greyscale panel, desaturate the image before you judge it. Two colours of similar brightness merge into one flat tone, so a picture that reads clearly in colour can lose its subject entirely in grey.
3. Set the black and white points. Expand the tonal range so the darkest area is genuinely black and the lightest is genuinely white. E-ink has a contrast ratio of roughly 15:1 to 20:1 against 1000:1 or more on a phone, so anything relying on subtle mid-tone separation flattens out.
4. Check for smooth gradients. Look for large areas of gradual tone — skies, studio backdrops, soft vignettes. These are what band on e-ink, because the panel shows 16 levels of grey rather than 256.
5. Dither to 16 grey levels yourself. If the image has gradients, quantize it to 16 evenly spaced greys with Floyd-Steinberg dithering before you export. Doing it yourself trades hard banding for fine noise that is invisible at reading distance on a 300 ppi panel.
6. Apply a light unsharp mask. E-ink scatters light rather than emitting it, which costs apparent edge definition. A small amount of sharpening compensates. Keep it subtle — halos around high-contrast edges are very visible against flat grey.
7. Export as PNG. PNG is lossless, so it carries no JPEG ringing around high-contrast edges. Those artefacts are usually invisible on a backlit screen but survive quantization to 16 greys and show up on e-ink.

Steps 1 and 7 are handled for you by the resizer on this site — pick your device, crop, and it exports at the exact panel resolution without uploading anything anywhere. Steps 2 to 6 are edits to make in your image editor before you bring the file here.

Troubleshooting by symptom

Does pixel density change any of this?

It changes two things: how much detail survives, and how visible dithering is.

At 300 ppi — most current Kindles, the Kobo Clara and Libra ranges, the Kobo Sage, the Boox Page, the PocketBook Era — thin linework and small type reproduce cleanly, and dither noise is genuinely invisible at reading distance. You can be ambitious with detail.

At 212 to 229 ppi — the Kobo Nia, the 10.3-inch note-taking devices and the reMarkable range — fine detail softens and dither noise starts to become perceptible as texture. Scale your detail up, use heavier line weights, and prefer bold composition over intricacy.

Note that pixel density has nothing to do with sizing. Only the pixel dimensions determine whether an image fits; density determines what it looks like once it does. The two get conflated constantly.

Frequently asked questions

How many shades of grey can e-ink display?

Sixteen. Standard E Ink Carta panels are 4-bit, giving 16 levels of grey rather than the 256 in an 8-bit image. Every tone in your picture is mapped onto one of those 16 steps, which is why smooth gradients such as skies band into visible stripes on an e-reader when they looked perfectly smooth on your monitor.

Why does my e-reader screensaver look washed out?

Because e-ink is reflective, not backlit. Its white is room light bouncing off a light-grey surface, and its contrast ratio is roughly 15:1 to 20:1 depending on panel generation, against 1000:1 or more on a phone. Mid-tones compress badly as a result. Expanding the black and white points in your source image before exporting recovers most of the apparent contrast.

Should I dither an image before putting it on an e-reader?

Only if it contains smooth gradients. The device quantizes to 16 greys either way — dithering just decides whether that happens as hard banding or as fine noise. Photographs with skies or soft backgrounds benefit clearly. Graphic images with flat areas and hard edges do not, and can look worse with dither noise added.

Is PNG or JPG better for an e-reader screensaver?

PNG. It is lossless, so it carries none of the ringing JPEG leaves around high-contrast edges — artefacts that are invisible on a backlit screen but survive the reduction to 16 grey levels and become visible on e-ink. File size is not a real constraint for a single screensaver, so there is nothing to gain from JPEG.

Do colour e-readers change how I should prepare an image?

Yes, and it depends which panel. Kaleido 3 models — Kindle Colorsoft, Kobo Libra Colour, Kobo Clara Colour — put a colour filter over the greyscale grid, so greyscale keeps the full 300 ppi while colour resolves at roughly 150. Flat areas of saturated colour hold up; fine coloured detail does not. Gallery 3, used in the reMarkable Paper Pro, keeps full resolution in colour instead.

Does pixel density change how I should prepare an image?

It changes how much detail survives, and how visible dither noise is. At 300 ppi, thin linework and small type reproduce cleanly and dithering is invisible at reading distance. At 212 to 229 ppi — the Kobo Nia, Kobo Elipsa 2E, reMarkable 2 and the 10.3-inch Boox models — fine detail softens and dither noise starts to be perceptible, so favour bolder shapes and heavier line weights.

Everything above assumes the image is already the right size, because nothing else helps if it is not. Start with the exact size for your device, then come back and work through the tonal steps — in that order, the difference is obvious on the first try.