turn a photo into cover art that feels printed, not prompted. halftone, riso, dither, grain, all in your browser. no models, no upload, ready for spotify and vinyl.
An album cover has to work at three sizes at once. It is a 3000 pixel square in the delivery spec, a fifty pixel thumbnail in a streaming queue, and, if the record gets pressed, a twelve inch object somebody holds. Most covers are designed for the first and fail at the other two. Designing for the thumbnail first, then adding detail that rewards the large version, is the practical order of work.
Streaming distributors converged on a shared minimum some years ago, and it has not moved much. A square JPEG or PNG, RGB, at least 3000 by 3000 pixels, under about ten megabytes. Spotify, Apple Music, Bandcamp and the aggregators that feed them all accept that, and most reject anything smaller than 1400 square.
Physical formats are a different problem. A twelve inch sleeve is 12.375 inches square with a bleed, which at 300 dpi is roughly 3712 pixels, and it needs CMYK rather than RGB. A CD booklet is 4.75 inches. Cassette J-cards are stranger still.
The consequential difference is colour space. A cover designed in RGB with saturated cyan or bright orange will shift noticeably when converted to CMYK for print, because those colours sit outside what four process inks can reproduce. If a physical run is likely, it is worth checking the design against a print-oriented palette early rather than discovering the shift at the proof stage.
At fifty pixels square, roughly the size a cover appears in a phone queue, almost nothing survives. What does survive is silhouette, one or two dominant colours, and the overall value structure.
Type is the first casualty. A band name set at what looks like a reasonable size on a 3000 pixel canvas is illegible below about two hundred pixels. Either set the type large enough that it survives the thumbnail, or accept that the thumbnail is purely a colour and shape signal and put the name where it works at full size.
Test properly. Export at 3000, then view the file at fifty pixels and at two hundred. If you cannot tell your cover apart from the three above and below it in a list, the design is not doing its job regardless of how it looks at full size.
Strong value contrast between subject and background does more work at small sizes than any amount of detail.
Album art has an unusually strong visual tradition, and most of it comes from printing constraints rather than from digital tools. Halftone dots from offset presses, riso ink overlaps from cheap short runs, photocopied textures from hardcore and punk sleeves, dithered images from early computer graphics, high contrast duotones from budget two colour printing.
Those looks carry meaning because they reference a physical production method and, by extension, a scene and an era. A halftone photograph in two inks reads as a small pressing. A crushed photocopy reads as a demo. A clean gradient reads as a major label campaign.
They also survive compression. Streaming services re-encode artwork aggressively, and a high frequency dithered or halftoned image holds up better than a smooth gradient, which bands visibly at low bitrates.
Bitgrain runs all of these as deterministic effects: fourteen dither algorithms, mono and colour and CMYK halftone, riso separation across nine inks, newsprint finish, duotone ramps, grain and cyanotype.
A cover is usually a layered document rather than a single processed image, which is why the studio matters as much as the effects. Photograph, a texture pass, a colour field, the artist name, the record title, sometimes a catalogue number and a small logo.
Keeping those as separate layers means you can re-export the same design as a square cover, a story-sized vertical for social, a banner, and a print-ready sleeve without rebuilding it. It also means you can run the texture treatment on the photograph without destroying the type.
Bitgrain's studio handles multi-layer documents with custom typography, a vector pen, design tokens for colour and type, and export to PNG, WebP, SVG and PDF at full resolution. Projects save as a portable .btg file that carries the layers, the effect settings and the palette together, so handing the file to a collaborator or reopening it next year gives the same result.
For a release campaign, the same document can drive the animated version too: the video timeline renders deterministic MP4 up to 4K, which covers the looping visual that most platforms now want alongside the static cover.
3000 by 3000 pixels, RGB, JPEG or PNG, under about ten megabytes covers every major streaming distributor. For a twelve inch vinyl sleeve you need roughly 3712 pixels square in CMYK to allow for bleed at 300 dpi.
Too much fine detail and type set too small. At fifty pixels only silhouette, dominant colour and value contrast survive. Design for that size first and add detail that rewards the large version.
RGB for streaming, which is most releases. If a physical pressing is likely, check the design against print colours early, because saturated RGB colours shift noticeably when converted to four process inks.
Better than smooth gradients do. Dithered and halftoned artwork is high frequency detail that re-encodes reasonably, whereas soft gradients band visibly at the bitrates services use for artwork.