Why audiobook covers are square
Print covers are portrait at 1.5 or 1.6. eBook covers are portrait at the same ratios. Audiobook covers are square at 1:1. The product is not the same rectangle at a different size; it is a different shape that is displayed in different contexts, and the art must be built for that shape.
Audio apps show the cover as a square in lists, grids, and the now-playing card. In a car, the square is the art next to the progress bar. On a phone, the square is the art in the library grid. On a desktop, the square is the art in the retailer's category page. A portrait eBook cover at 1600 by 2560 placed in a square frame will be pillarboxed with bars or center-cropped through the title. A square at 3000 by 3000 placed in a portrait frame will be letterboxed. The square is not a resized eBook cover. It is a separate master built from the same layered file but re-set for the square.
The tool checks the three things that determine whether the square will be accepted and whether it will look sharp: is it square, does it meet the minimum, and is it at the recommended size. A file can pass the minimum and still be soft, because the minimum is what the uploader will accept before rejecting, not what looks sharp on a high-density display. The recommended is what looks sharp.
The spec that actually matters
Most audiobook retailers converge on the same numbers, with small differences in the file cap. The recommended is 3000 by 3000 pixels, 1:1, in sRGB, as a baseline JPEG, at 300 DPI, under the retailer's file cap. The minimum is 2400 on the short side for ACX and Findaway and many others; Kobo Audio accepts 2000 as the minimum. The maximum file size is 50MB for ACX, Findaway, Apple, and Google; Kobo enforces 5MB, which is the tightest, and Spotify via Findaway Open Access enforces 20MB.
The numbers are not arbitrary. At 3000 by 3000 the cover is 9.0 megapixels. At 2400 it is 5.76 megapixels, or about two thirds of the detail. On a high-density phone the 3000 file downscales with a light unsharp mask and remains crisp; the 2400 file upscaled to 3000 invents pixels without inventing detail and looks soft next to native 3000 comps. The 300 DPI tag does not add detail beyond the pixels, but it signals that the file was exported at print-quality sharpness at those pixels, not as a small file tagged 300 DPI.
Color is sRGB for every audio retailer. A master built in Adobe RGB or Display P3 will be converted to sRGB by the pipeline and the shift is most visible in saturated reds and oranges. Convert to sRGB on export and preview the conversion. The file format is baseline JPEG, not progressive, at quality 90 to 95. A 3000 by 3000 sRGB baseline JPEG at that quality is typically 3 to 5 MB, well under the lowest cap. A TIFF at the same pixels is 27 MB and offers no visible advantage on screen. Save JPEG, keep the layered master as a PSD or Affinity file, and keep the JPEG under the smallest cap you publish to so the same file passes everywhere.
Platform differences that change the decision
ACX and Audible are the default for most authors because they feed the largest audio audience. The spec is 3000 by 3000, 1:1, 2400 minimum, sRGB, baseline JPEG, under 50MB. The compliance that ACX actually enforces beyond the numbers is no blur, no pixelation, and no placeholder art, and title and author on the cover must match the metadata. A cover that is square and 3000 but has a blurry, upscaled image will be rejected for quality, not for pixels.
Findaway Voices distributes to forty plus retailers and uses the square as the common denominator. The spec is the same 3000 by 3000, 2400 minimum, sRGB, JPEG, under 50MB. The note the tool shows for Findaway is to keep type inside a 10% inset from the edge, because the square crops less forgivingly at the corners where some retailers round the thumbnail. Type that sits 5% from the edge on an eBook portrait may sit 5% from the edge on a square as well, but the square's corners are more visible in the app, so keep the inset.
Apple Books Audio also uses 3000 by 3000, 2400 minimum, sRGB, JPEG, under 50MB, but Apple's library shows the square at a slightly smaller rendered size where tiny type vanishes. A title set at 12 points on a 3000 square may be readable at full size and not at 200 pixels in the library grid. Keep the title large and the author at roughly forty to fifty-five percent of the title size, just as on the eBook, but test the square at 200 pixels wide, not just at 3000.
Kobo Audio is the strictest on file size. The spec is still 3000 by 3000, but the cap is 5MB. A 3000 square saved at quality 98 or with embedded metadata can be 6 to 8 MB and will be rejected for size, not for pixels. Lower quality to 90 and strip metadata; the cover will still look sharp at 200 pixels and will pass the cap.
Google Play Audio is also 3000 by 3000, 2400 minimum, sRGB, JPEG, under 50MB, but Google re-encodes more aggressively than the others. A sharp master at 3000 survives the store's compression better than a 2400 master. Start sharp.
Spotify via Findaway Open Access is 3000 by 3000, 2000 minimum, sRGB, JPEG, under 20MB. The square is shown small in lists and as the now-playing art. Keep the title legible at 120 pixels, not just at 3000, because that is the size the listener sees while the audio plays.
Worked examples with the compliance shown
Note: the following examples are hypothetical and illustrative rather than a sourced case study.
A cover exported at 3000 by 3000 for ACX is the ideal that passes everywhere. Width 3000, height 3000, square, 9.0 megapixels, at the recommended, above the 2400 minimum, 1:1, sRGB, baseline JPEG, under 50MB. The verdict is meets spec — ready to upload. The metrics show your file 3000 by 3000, recommended 3000 by 3000, minimum 2400 by 2400, ratio square. The checklist shows square checked, meets minimum checked, at recommended checked, and sRGB JPEG under cap checked. The breakdown lists retailer ACX/Audible, width 3000, height 3000, ratio 1:1 square, min 2400, DPI 300, format JPEG or TIFF, max 50MB, color sRGB, and the note about square and title matching metadata. The recommendations say to keep the square master at 3000 and derive smaller squares from it, not to upscale a smaller file.
A cover at 2400 by 2400 for Findaway meets the minimum but not the recommended. Width 2400, height 2400, square, 5.76 megapixels, above the 2400 minimum but below the 3000 recommended. The verdict is passes minimum but not recommended — softer on retina. The metrics show your file 2400 by 2400, recommended 3000 by 3000, minimum 2400, ratio square. The checklist shows square checked, meets minimum checked, at recommended unchecked. The breakdown and table are the same, but the recommendations say to re-export from the layered master at 3000 with a light sharpen, not to upscale the 2400 file, and to keep the file under the smallest cap you publish to by lowering JPEG quality to 90 rather than shrinking pixels.
A portrait eBook cover reused for audio at 1600 by 2560 is the common fail. Width 1600, height 2560, rectangular portrait, 4.1 megapixels, above the minimum in pixels but not square, ratio 1:1.6. The verdict is not square — audio requires 1:1. The body says the cover will be rejected or auto-cropped, often through the title. The metrics show your file 1600 by 2560, recommended 3000 by 3000, minimum 2400, ratio portrait. The checklist shows square unchecked, meets minimum unchecked in the square sense, at recommended unchecked. The recommendations say to rebuild the cover as square, not to stretch the portrait to square, and to re-set the title for the square and re-center the art. Stretching the portrait makes the art wider and the type wider, and the title that looked confident at 1.6 looks condensed at 1:1.
Color, type, and the square's own hierarchy
The square has its own hierarchy that is not just the portrait scaled. The portrait title may have been set at a size where the longest line filled seventy to eighty percent of the trim width. The square title should fill a similar proportion of the square width, but the square is wider relative to its height, so the title may want to be slightly larger or to break differently. A two-line portrait title may become a one-line square title, and a one-line portrait title may need to break into two lines on the square to keep the point size large enough to read at 200 pixels.
Color the square for the square, not for the portrait. A portrait with a dark top and a light bottom may have the title on the dark top. The square cropped from the center may place that dark top in the middle, and the title may lose its contrast. Re-evaluate the title color for the square: white on dark or dark on light at high contrast, not mid-tone on mid-tone. The square's background behind the title should be the quietest area of the image, even if that area is not where the portrait's title sat.
Type the square for the square. The portrait's author size at forty to fifty-five percent of the title size may be too small on the square where the title is larger relative to the canvas. Keep the author at the same percentage, but check the result at 200 pixels wide. If the author vanishes, increase the author size slightly or increase the tracking. The body on the square is still the quieter family, but quiet does not mean invisible.
Edge cases: portrait masters, old files, and type at small size
A portrait master at 6000 by 9600 cannot be exported as a square by changing the canvas size alone. The portrait must be re-composed for the square. Place the art centered, keep the title large, and keep type inside a 10% inset from the edge. The square's corners are more visible than the portrait's corners, because audio apps often round the thumbnail. Type that sits 5% from the edge on a portrait may sit too close on a square where the corner radius clips. Keep the inset.
Old files at 2400 or 2000 that were once the recommended are now the minimum. The recommended has moved to 3000 as device density increased. A 2400 file that was the ideal three years ago will still pass upload, but it will be softer on high-density displays than a native 3000 file. Do not batch-upscale old covers. Re-export from the layered masters at 3000 with a light unsharp mask. The recompression will look sharper than an upscaled JPEG.
Type at small size is the hidden edge case. An audio cover is often seen at 200 pixels wide in a library grid and at 120 pixels in now-playing. A title set at 12 points on a 3000 square may be readable at full size and not at 200. Keep the title large — roughly the same size relative to the square as the eBook title is to the portrait — and keep the author at forty to fifty-five percent of the title size in the body font. A subtitle that is expected to read at full size on an eBook is not expected to read at 200 pixels on audio; keep the square title to the main title and let the subtitle live in the metadata.
File size versus pixels is the other edge case. A 3000 square at quality 98 may be 7 MB and will pass ACX but fail Kobo's 5 MB cap. The fix is not to shrink pixels to 2400, but to lower JPEG quality to 90 and strip metadata. The cover at 3000 at quality 90 is typically under 4 MB and still sharp at 200 pixels, while a 2400 cover at quality 98 is softer at 200 pixels and still over 5 MB if the metadata is large.
How to build the square after you have the numbers
Open the layered master at 4000 to 6000 on the long side and create a new square canvas at 3000 by 3000. Do not stretch the portrait. Place the art centered and scale it to fill the square while keeping the focal point centered. If the art is a figure, keep the face centered; if it is a landscape, keep the horizon at the same relative height as on the portrait but crop the sides. The square should feel like the same cover seen through a different window, not like the portrait squeezed.
Set the title anew for the square. The portrait title may have been set at a size where the longest line filled seventy to eighty percent of the trim width. The square title should fill a similar proportion of the square width, but the line breaks may differ. A two-line portrait title may become a one-line square title or vice versa. Re-break the title for the square at a natural phrase, keep leading generous, and keep all type inside a 10% inset from the edge. Do not reuse the portrait type block by scaling it to fit the square; re-set it.
Set the author and any series line in the body font at roughly forty to fifty-five percent of the title size, centered, with at least 0.05 inches of quiet on each side at print-equivalent. The body on the square should be the quieter of the two families, just as on the portrait.
Export as sRGB baseline JPEG, quality 90 to 95, at exactly 3000 by 3000. Check the byte size in the file manager before upload, not just the pixel dimensions in the design tool. A 3000 square at quality 92 should be well under 5 MB; at quality 98 it may exceed Kobo's cap. Lower quality to 90 rather than shrinking pixels.
Keep the layered square master as a separate file from the portrait master. When the next book in the series needs a square, duplicate the square master, not the portrait, and swap the art. The square system will then stay consistent, just as the portrait series system does.
How to test the square for real
Export at 3000 by 3000 and open it at 100% in a viewer that does not smooth. At 100% the image is larger than most displays, so zoom to the size the store actually shows — roughly 200 to 300 pixels wide in a library grid and 120 pixels in now-playing — and see whether the title reads without leaning in. Then upload to the retailer's previewer: ACX has a cover preview that shows the square in the product page context, Findaway shows the square in the distribution preview, and Apple shows the square in the Books preview. The preview shows the re-encoded result, which is the ground truth.
Test on a phone. Open the live listing in an incognito window after publication and view the library grid at default zoom, then as a customer who searches a category that surfaces your ASIN or a close comp. The square on a phone at arm's length is the smallest context the cover must survive. If the title reads there, it reads everywhere it matters. If it does not, the fix is not more pixels but larger type, higher contrast, or a cleaner background behind the letters, as the thumbnail legibility tool measures.
Test the file size as well as the pixels. A file that is exactly 3000 by 3000 but 7 MB will pass ACX and fail Kobo. The tool reports the max size per retailer precisely for that reason. Check the exported JPEG's byte size before upload.
Common mistakes and how to avoid them
Stretching a portrait eBook cover to square is the most common. The portrait at 1600 by 2560 stretched to 3000 by 3000 makes the art wider and the type wider, and the title that looked confident now looks condensed. Re-set the title for the square from the layered master, not by transforming the flat JPEG.
Exporting at the minimum rather than the recommended is the second. The uploader accepts 2400, but the product page and the thumbnail are built from that minimum, so the cover is soft next to comps that exported at 3000. Export at the recommended every time; the minimum is a floor, not a target.
Adding bleed to an audio cover is the third. Bleed is for print wrap covers where the trimmer removes excess. Audio covers are displayed edge to edge without trim, so the file should be exactly the square pixels without extra bleed edges. A 3000 square with an added 0.125-inch bleed built at pixels will be the wrong ratio and will be padded with bars.
Uploading in the wrong color profile is the fourth. A cover saved as CMYK or Display P3 will be converted to sRGB by the pipeline, and the conversion shifts saturated colors. Convert to sRGB on export and preview the conversion. The shift is most visible in reds and oranges.
Saving as PNG when JPEG is expected is the fifth. PNG at 3000 by 3000 is 27 MB without visible advantage; some retailers re-encode PNG more heavily than JPEG and the result can look softer. Save JPEG baseline at quality 90 to 95 and keep the PNG master only as an archival layered file.
Forgetting the square inset is the sixth. Type that sits 5% from the edge on a portrait may sit too close on a square where the corners are more visible and some apps round the thumbnail. Keep type inside a 10% inset from the edge on all sides for the square, even though the portrait uses 5%. The square's corners are the risk.
Reusing a low-resolution file by upscaling is the seventh. A 1400 file upscaled to 3000 invents pixels without inventing detail and produces a soft cover that still fails the sharpness test. Re-export from the layered master at 3000 with a light unsharp mask.
Final checks before you commit and upload
Run the tool once with the platform you actually publish to and once with a different platform where the file cap is stricter — for example, ACX at 50MB versus Kobo at 5MB. If the same file must serve both, see whether the file size is under the stricter cap. If it is not, lower JPEG quality to 90 and re-export rather than shrinking pixels.
Open the exported square at 3000 by 3000 and check the title distance from the edge. It should be at least 10% inside the frame on all sides, so a small crop or corner rounding does not clip. Do not add bleed; keep the file at the exact square pixels without extra edges.
Check the byte size. A 3000 square at quality 92 should be well under 5 MB; at quality 98 it may exceed Kobo's cap. Lower quality to 90 and re-export rather than resizing pixels down, because pixels determine sharpness and quality determines file size more than the last 100 pixels do.
Upload and verify live. In the retailer's previewer, confirm the cover fills the frame without bars or softness and that the title reads at 200 pixels wide. On a phone listing in an incognito window, confirm the square sits with the comps at arm's length. If it does, lock the export settings — pixels, ratio, profile, quality, and file size — alongside the KDP file name. When the next book grows, reuse the same settings rather than relying on memory that the last cover was 3000. The numbers are fixed; your file is the variable.