A gradient in a deck is a structural decision, not surface polish. It sets which text colour the slide can carry, it decides where the eye enters the frame, and it does both before a word is written. Gradient backgrounds look cheap when that decision is made last, dropped behind finished content as a substitute for having designed the slide. They look deliberate when it is made first.
Why gradient backgrounds usually look cheap
The cheapness is never the gradient itself. It is the ordering. A deck gets built in white, the content gets placed against a white assumption, and somebody then opens a background picker and pulls a two-stop violet-to-blue ramp across all forty slides. Nothing on those slides was designed for it, so the gradient works as compensation rather than structure — and compensation is what the eye picks up.
You can watch the failure finish itself off in the fix people reach for next. The text is now unreadable across half the frame, so a black overlay goes on at forty per cent. That overlay flattens the ramp into an even murk: the wash chosen for its colour is now grey, with a hint of the colour surviving in two corners. Two decisions cancelling each other out.
Applying the same wash to every slide is the other half of it. A background appearing forty times running is not emphasis, it is wallpaper. The decks worth copying use an atmospheric background on perhaps six slides out of forty: the cover, the section breaks, and the number the argument turns on.
What a wash decides before you have written a word
In Slideable a background gradient is called a wash, and every wash carries a dark flag. That flag is what tells slide text to default to white. One boolean for the whole gradient sounds like a simplification until you notice what it forces: choosing a wash is choosing the text colour for every slide that uses it, while the slide is still empty.
The second thing a wash decides is where the reader enters the frame. A radial wash resolves to a wide ellipse — 120% by 120% at a focal point — because anything tighter reads as a bullseye at 16:9 rather than as light arriving from somewhere. That bright region is read first, ahead of the type. Leave it centred under a title sitting in the upper third and the title becomes the second thing read: a composition you failed to notice rather than chose.
- Text colour: a wash you cannot commit to one text colour is a wash you cannot run full-bleed.
- Entry point: the brightest region of the ramp is read before any type sitting on it.
- Weight: a wash holding dark across the top third leaves a title somewhere to live; one peaking in the centre does not.
This is also why a Position control can exist at all. Because a wash is stored as a recipe rather than a rendered picture, moving its focal point re-projects it: linear washes rotate, radial washes move their centre. The gradient is recomputed, never cropped or stretched, so changing where the eye enters costs nothing and stays reversible.
Why gradients band, and what stop positions actually fix
Banding is arithmetic, not bad luck. A screen gives you 256 levels per channel. If a gradient crosses twenty levels of red over a 1920-pixel slide, each level occupies 96 pixels, and 96-pixel stripes are not subtle. The working threshold: bands separate once a single colour level covers more than about four pixels.
That is why dark washes band worst. Near-black gradients cross very few levels — the carbon wash runs #08080a to #26262c, thirty steps of red across the whole slide — where a pale one crosses hundreds. It is also why banding turns up in the room and not at your desk: an eight-bit projector link, a compressed screen share, and the background that looked smooth on your laptop arrives striped.
Two fixes get suggested and one works. Adding stops does nothing if the new ones sit on the straight line between the old ones; you have described the same ramp using more words. What works is stops that bend the path through colour space, so the ramp crosses more distinct values, and positions that spend screen distance where the colour is actually moving.
That is what the optional positions on a wash are for. Without them the stops spread evenly, which is right for a soft blend and wrong for a wash that has to hold one colour before it turns. The night wash is five stops at 0, 28, 58, 84 and 100 per cent: black holds through the top third, the ramp does its real work in the two middle segments, and the last sixteen per cent carries the lightest move. The flat part is flat deliberately, so nothing there can band; the part that moves covers its distance fast enough that the steps stop separating.
Why a gradient behind body text is nearly always wrong
Contrast against a gradient is not a number, it is a range. Take the dawn wash: sand at #f6d5a8 running to violet at #8f7bb5. Black text is comfortable at the sand end and marginal at the violet end, and no single colour is right across the whole ramp. The dark flag promises that one text colour works everywhere on that wash, and most washes cannot honestly make that promise to a full column of body copy.
A title dodges this, because a title is six words you can place in the region you picked. Eleven lines of body text cannot: the block spans the ramp by definition, and somewhere in the third paragraph the contrast ratio drops under 4.5:1 and stays there. Nobody in the audience thinks about contrast ratios. They stop reading, and you lose the paragraph you spent an hour on.
A background you have to darken with an overlay before the text works is a background you did not choose. You chose the overlay.
So the rule is not that gradients are banned. Washes belong on the slides carrying fewer than roughly ten words — covers, section breaks, the single statement, the one number — and every slide with a paragraph gets a flat surface. Keep that flat surface in the same colour family as the wash and the deck still reads as one object.
Why photographed weather reads as intentional and an even blend does not
An even three-stop blend has no subject. Every point on it is transitional, on its way from one colour to another, so nowhere on the slide is the background being anything in particular. It reads as a default because it is one: precisely what you get when you supply three colours and no positions.
Weather does not behave that way. A sky holds one colour through most of its height and turns near the horizon, and the held part is what makes the turn mean anything. The harbour wash has that shape: #0c1a2b sitting through the first thirty-four per cent before it opens out to #7c96b4. It reads as one particular evening rather than a slider somebody dragged.
The angle does the same job at a smaller scale. Harbour runs at 170 degrees, carbon at 175, the library default at 160 — all off the vertical and well off the obvious 45. Ten degrees off vertical is not a look anyone consciously notices. It is the difference between a background that was set and one that was accepted.
Why storing a recipe keeps it a decision you can change
A wash in Slideable is stored as a colour recipe: a list of stops, optional stop positions, an angle, and the dark flag. Not a CSS string, and emphatically not an image. That storage choice is what makes every argument above survivable past the first draft.
A gradient shipped as a picture is an asset. Retune it and you re-export it, then hunt for the slides still carrying the stale copy — and by slide forty nobody hunts. The gradient has stopped being a decision and become something the deck inherited. A recipe has no stale copies. Change a stop and every surface using that wash changes with it, including the ones you had forgotten.
The library is shared as well: this site and the editor read the same file. The gradient behind this article is the entry the Background panel offers inside the product, resolved at render time rather than exported as a thumbnail. A wash retuned in one place is retuned in both, and the site cannot advertise a gradient the editor cannot produce — a stricter constraint than most marketing pages accept.
None of this is a large feature. It sits underneath 106 layout archetypes, eight chart types and an MCP server exposing 52 tools, all downstream of one idea: the decisions worth automating are the ones made before the content arrives, not the polish applied after. Open a deck in the browser — no account — and set the wash first. Then write the title, already knowing what colour it is.
Common questions
- Why do gradient backgrounds look cheap in presentations?
- They look cheap when the gradient is applied at the end, over content that was designed against a white background. The gradient then has to compensate for a layout that ignores it, which usually means adding a dark overlay that flattens the gradient into murk. Choosing the gradient first, so it determines text colour and where the eye enters the slide, removes the problem at its source.
- How do I stop a gradient from banding?
- Banding happens because a screen has only 256 levels per colour channel, so a gradient crossing twenty levels over a 1920-pixel slide shows a visible step every 96 pixels. Adding more stops along the same straight line does not help; what helps is intermediate stops that bend the path through colour space and stop positions that compress the flat part of the ramp so the moving part covers its distance quickly. Dark, near-black gradients band worst because they traverse the fewest levels, and projectors over eight-bit links make it worse again.
- Should I put body text on a gradient background?
- Almost never. Contrast against a gradient is a range rather than a single value, so a full-height text block will start above 4.5:1 at one end of the ramp and fall below it at the other, with no text colour that works across the whole slide. Reserve gradients for slides carrying fewer than about ten words — covers, section breaks, a single statement or number — and give anything with a paragraph on it a flat background in the same colour family.