Why an image can change color after export—and how to investigate

Compare the same file under controlled conditions

When an exported image looks different, begin by identifying exactly what is being compared. The editor may show the working document, while a browser shows a compressed export on a different background. The device may also have brightness or display adjustments enabled. Change one variable at a time. Open the same exported file in the editor and in the intended viewer before deciding that the export itself changed the colors.

Keep a copy of the original project and record the export settings. Do not repeatedly change profiles until something looks pleasing on one screen. That can create a file that compensates for one viewing condition while becoming less predictable elsewhere. The first task is diagnosis: determine whether the difference comes from the document, the export, the viewer or the display environment. A controlled comparison is more useful than a sequence of adjustments whose effects cannot be traced.

Understand what a profile contributes

A color profile supplies context for interpreting color values. The same numerical values do not necessarily describe the same appearance in every color space. Krita's official color-management documentation explains the relationship between image spaces, display management and export choices. You do not need to master every part of color science to recognize that a profile is meaningful information rather than decorative metadata.

For an ordinary web image, use a workflow suited to the destination and keep the profile handling consistent. For print or specialist production, follow the recipient's specifications. Avoid replacing a profile solely because its name is unfamiliar. Assigning a different interpretation and converting an image for a new destination are not interchangeable concepts. If you are unsure which operation a command performs, consult the editor's documentation before applying it to the only working copy. Preserve the source so that a mistaken conversion can be undone by returning to known material.

Separate soft proofing from the final image

Some creative tools can simulate a target output condition for preview. That preview is useful, but it is not necessarily identical to the underlying image or to every exported file. If a proofing view is active, note that before comparing screenshots or deciding that the colors shifted. The viewer may be showing an intentional simulation of a printer or another output condition rather than a neutral display of the document.

Use a small diagnostic image or a copy of your project to understand the current settings. Include neutral tones, saturated colors and a smooth gradient so that differences are easier to see. Do not treat the diagnostic image as a calibrated scientific test unless you have the equipment and process to support that claim. Its purpose is practical troubleshooting: reveal whether a setting produces a systematic change and help you avoid changing several unrelated controls at once.

Check the export's actual properties

Inspect the exported file's dimensions, format and profile information where your tools expose them. A file may have been resized, converted or saved with different settings than expected. Confirm that you opened the newest export rather than an older file with a similar name. This sounds simple, but repeated exports to several folders can make an apparent color-management problem partly a version-selection problem.

Return to the source and create one clearly named test export with documented settings. Compare it with the previous export. If the change is visible only in one viewer, investigate that viewer's handling rather than continuing to alter the source image. If it is visible in every viewer, inspect the export path more closely. This sequence narrows the cause. It also preserves a record of what changed, which is useful when a collaborator needs to reproduce or diagnose the issue on another machine.

Keep the two test exports visible under distinct names and note the single setting that differs. If you cannot describe that difference in one sentence, the comparison probably changes too many variables to isolate the cause reliably.

Judge the image in its real context

The background surrounding an image affects perceived contrast and color. A photograph on a dark editing canvas may feel different on a white webpage even when the file is unchanged. View the image in the actual layout or a close preview. Check whether the concern is a technical mismatch or an ordinary perceptual difference caused by context. These call for different solutions: correcting profile handling is not the same as adjusting a composition for its final page.

Use more than one representative device when the image will reach a broad audience, while accepting that displays vary. Do not chase exact visual identity across every uncalibrated screen. Focus on avoiding obvious errors and maintaining a sensible, documented output workflow. If precise reproduction is a contractual or production requirement, use the appropriate professional proofing process. A casual browser comparison cannot establish that level of certainty, and a single attractive screenshot is not evidence that the whole workflow is controlled.

Keep a reproducible export recipe

Once the result is understood, save a short recipe containing the source color setup, destination, export format and relevant profile choices. Include any proofing condition that affected the review. This turns a one-time fix into a repeatable process for the next asset. Keep the editable master separate from the destination-specific export so that future requirements do not force you to work backward from a converted file.

The useful outcome is not a claim that colors will look identical everywhere. It is a clear explanation of how the file was prepared and why the chosen settings fit its destination. When a difference appears later, you can compare against that known process. Color troubleshooting becomes manageable when you preserve the source, control the comparison and isolate each stage. Randomly changing settings until one preview looks right may hide the symptom, but it leaves the next export just as uncertain.