PPI vs DPI: Print-Size Test for AI Portraits

A portrait that looks sharp on a phone can disappoint on paper. The reason is rarely a single “DPI setting.” Its pixel dimensions, intended print size, viewing distance, printer process, and any enlargement all matter. This PPI vs DPI worksheet helps a creator work out what the file actually supports before promising a client a poster.

Test a licensed portrait with APOB Image Upscale.

The calculations below are worked examples, not results from a completed APOB print test. The production experiment needs a rights-cleared original, the downloaded upscale, a contact sheet, and a proof from the selected printer. Keep those artifacts together before claiming any print size is accepted.

Decode image pixels and printer dots before changing a file

The image and the printer have different units. Adobe's printed-image resolution guide describes how image pixels relate to a physical output size; its printer-resolution guide treats the printer's dots as a separate device property. A metadata box labeled “DPI” does not manufacture image detail. First read the pixel count, then decide the physical size, and only then discuss device output with the print shop.

File pixel dimensions

Open the untouched image in an editor or file-inspection utility and record width × height in pixels. Do this before resampling, changing metadata, exporting a derivative, or uploading to another tool. Write source_01 | 2400 × 3000 px | original file preserved rather than high resolution. The first entry is auditable; the second is a judgment without a target. Record the file format and orientation as well. A vertical portrait cropped to a square will have fewer usable pixels on one axis even when its underlying file is large.

For an APOB-generated image, inspect the download rather than assuming the preview has identical dimensions. Keep its permission note.

Printed physical size

Image PPI means the number of image pixels assigned per printed inch. Divide usable pixels on each axis by the chosen PPI to estimate the physical dimensions. A 2400 × 3000 px file placed at 300 PPI gives 8 × 10 inches before cropping. Set that same file to 12 × 15 inches and it has 200 pixels per inch at that size; renaming its metadata to “300 DPI” does not change those 2400 × 3000 pixels.

Ask the shop whether this file at the trimmed size and viewing distance can make an acceptable proof. Calculate from pixels remaining after crop.

Printer resolution

Printer DPI describes how a device places dots of ink or toner. It is not interchangeable with the source image's PPI. A device may place multiple dots to reproduce the color of one image pixel. Do not multiply a printer's marketed DPI by the image width to invent a required image size. Ask the supplier for its recommended input dimensions, color space/profile, bleed, sharpening policy, and proofing method. Adobe's print-resolution specifications explain why viewing distance affects the useful image resolution; 300 PPI is a common close-viewing reference, not a universal rule for every sign, printer, or substrate.

Calculate a target print from actual pixel counts

The worksheet is deliberately simple enough to use before any creative upscaling. It answers whether the original file already meets a target, whether a larger file is needed, and which dimension is the bottleneck. It does not predict perceived quality, because noise, focus, generated texture, paper, and lighting still affect the proof.

Source file

Record usable_width_px, usable_height_px, trim_width_in, and trim_height_in. The effective PPI on each axis is usable_pixels ÷ trim_inches; the lower of the two is the bottleneck. If the design has a 0.125-inch bleed on each side, calculate the supplied canvas separately from the final trim. If the portrait must fill a different aspect ratio, crop a duplicate first and measure the crop. Never overwrite the master merely to make the worksheet look cleaner.

Keep pre- and post-crop dimensions in the ledger. Pixel count is a gate, not proof of believable texture.

Viewing distance

A small framed portrait invites close inspection; a display viewed across a room does not. For the first, use a conservative close-viewing reference such as 300 PPI when estimating the starting target, then ask the printer whether its process and viewing conditions call for something else. For a large installation, the supplier may accept a lower effective PPI, but no single number applies to every job. Record the expected distance and whether a customer can walk up to the print.

When placement is unknown, calculate more than one PPI and label each assumption. Only the supplier can approve an alternative.

Calculation table

These are arithmetic examples, not APOB measurements or printer approvals.

For any row, print_width = pixel_width ÷ target_PPI. To estimate a requested source size, reverse it: required_width_px = print_width_in × target_PPI. A 12 × 15-inch close-viewing target at the illustrative 300 PPI calls for 3600 × 4500 usable pixels. A 2400 × 3000 source is therefore 1.5× short on both axes. That ratio is an enlargement goal, not evidence that a model can reconstruct authentic missing detail. Adobe's pixel and physical-size explanation is the reference for the unit conversion.

Upscale one licensed portrait for a larger print

This is the experiment to run before a customer order. The actual source, generated file, and proof are pending for this article; no before/after quality claim follows. The benefit of a controlled test is that the original, target, and defects remain identifiable when a retake is needed.

Source permission

Choose a portrait whose creator and subject permit editing, generation, and the intended print distribution. An original APOB portrait is one possible source, subject to its actual rights. Keep the license or signed release with the file. Avoid an unlicensed celebrity image or a customer's private photo without permission. Assign a source ID, preserve the original, and record its dimensions and checksum. Put any limited print license on the handoff card.

The APOB upscale feature can be a candidate production step, but its existence does not establish suitability for a particular print job. Use the same original for any comparison, and preserve every derivative as a separate file. A reader should be able to identify which output came from which setting without trying to infer it from appearance.

Target dimensions

Set the target from the calculation table: for example, a 12 × 15-inch trim at an illustrative 300 PPI would need 3600 × 4500 usable pixels. Add the printer's bleed and crop rules before final export. Record the selected model or upscale setting, upload time, output pixel dimensions, and file format. If APOB exports a size other than requested, the actual downloaded size is what belongs in the result table.

The contact sheet should show matched 100% crops from the actual downloads, with filenames and settings. Do not compare a recompressed messaging-app copy.

APOB prompt

Use a constraint-led instruction where the tool permits text guidance: preserve subject identity, expression, clothing details, background geometry, and any real text; improve apparent clarity without adding jewelry, changing eye shape, inventing logos, or smoothing away characteristic skin features. This is a proposed prompt, not a claim about how a specific APOB model responded. Record the exact prompt and any strength controls actually exposed by the interface. If no prompt field exists in the selected route, record the available controls instead of inventing one.

Keep rejected attempts in the log. Flag altered identity cues, seams, hands, and lettering even when the result looks sharper.

Inspect invented texture at final viewing distance

An upscale can add pixels, but new pixels are not necessarily true source information. The review therefore has two passes: inspect matching crops on screen and inspect a physical proof at the intended distance. Until both exist, the outcome remains “unverified.” The asset ledger should name the reviewer and decision rather than claiming generic “enhanced quality.”

100% crops

Export matched crops of eyes, hair edge, skin, fabric, hands, and any text or logos. Label the original and each enlarged output with dimensions and attempt number. At 100% display magnification, mark duplicated strands, waxy skin, shifted seams, altered lettering, halos, or unexpected objects. A side-by-side crop is not a scientific benchmark by itself; it makes the defect visible enough for a reviewer to decide whether another route is needed.

Compare lettering character by character. A changed customer name or identity cue is a rejection, not an acceptable enhancement.

Print proof

Send the actual export to the intended printer, with trim dimensions, bleed, color-profile instructions, and the declared effective PPI. Request one physical proof on the planned stock and finish. View it first at the expected distance, then close enough to inspect face and text. Record lighting and proof date. A screen proof can reveal geometry and some tonal problems, but it does not substitute for a physical proof when the output medium matters.

If the printer requests a different profile or format, make a new derivative. Adobe's guidance does not replace that shop's specifications.

Accept/reject marks

Use three states: ACCEPT when the crop and proof meet the signed brief, REVISE when a bounded defect may be corrected, and REJECT when identity, text, permission, or print integrity fails. A row might read hair edge | REVISE | duplicated strand at 100% | reviewer/date; that is an example of a log entry, not an observed result from this article. Add a link to the crop and proof image when the real test is performed. No evidence link means no pass claim.

Deliver a file a print shop can evaluate

The handoff is a small production card, not a marketing claim. It travels with the image so a new printer or designer can recreate the assumptions. It should distinguish verified measurements from still-open choices and specify where the original and proof live. This is more useful than emailing “300 DPI attached” without dimensions.

File name

Use portrait_[asset-id]_[trim-size]_[output-px]_v03.tif as a naming template; preserve the original separately and keep customer names out of public URLs.

Dimensions

Write the actual width × height in pixels, trim size, bleed, effective PPI on both axes, crop note, format, embedded profile, and viewing distance. If the printer has supplied a separate specification, attach its date and contact. Mark any value derived from an estimate. A change in trim size after approval invalidates the previous effective-PPI calculation and may require a new proof.

Proof request

Ask the supplier: “Please confirm whether this file, at the stated trim size and stock, meets your input requirements; advise on bleed, profile, format, sharpening, and a physical proof.” Provide the review crops and disclose AI enlargement where that matters to the job. Do not claim that APOB recovered lost source detail. The final acceptance belongs to the buyer and printer after they see the actual file and proof. The next step is a small, rights-cleared test, not an unsupported guarantee.

Sources

• Adobe, Printed Image Resolution

• Adobe, Resolution Specifications for Printing Images

• Adobe, Printer Resolution

• APOB, AI Image Upscale