{"course":{"slug":"image-editing","title":"Editing an Image","url":"https://addaly.com/learn/image-editing","level":"beginner","track":"create"},"lesson":{"slug":"skin-tones-honestly","title":"There is no correct number for skin","url":"https://addaly.com/learn/image-editing/skin-tones-honestly","position":38,"of":78,"minutes":10,"answer":"Skin has no universal correct value, so the method is to neutralise from a known reference, check hue consistency across the face rather than against a target.","keyIdea":"Skin has no universal correct value, so the method is to neutralise from a known reference, check hue consistency across the face rather than against a target, and recognise that automatic exposure and white balance were tuned on distributions that systematically under-serve darker skin."},"body":{"format":"markdown","text":"## The rule you will be told, and its actual status\n\nSearch for skin tone correction and you will find a rule: in CMYK, yellow should be slightly higher than magenta, and cyan should be a fraction of magenta: often quoted as cyan being a quarter to a third of the magenta value, with magenta and yellow close together.\n\nIt is a genuine rule of thumb from the pre-press era, and it works within its range: it was derived from Caucasian skin printed on coated stock. Applied to darker skin, the ratios change substantially: cyan rises considerably relative to magenta, and the yellow-magenta relationship shifts. Applied as a universal target, it lightens and pinkens dark skin, which is exactly the failure mode this lesson is about.\n\nThe honest statement: **skin covers an enormous range of hues, lightnesses and saturations, and there is no single set of numbers that is correct.** What there is instead is a method.\n\n## The method that works\n\n**1. Get neutral first.** Set white balance from something known to be neutral in the frame, as in the earlier lesson. Most \"the skin looks wrong\" problems are white balance problems, and fixing skin directly while the whole frame has a cast means fighting the same error twice.\n\n**2. Check that the cast is gone from something that is not skin.** A white shirt, a grey wall, the whites of the eyes. If those are neutral and the skin still looks wrong, the problem is in the skin.\n\n**3. Look for hue consistency across the face rather than a target value.** Skin should be one hue family across the whole face, varying in lightness and saturation. The commonest genuine fault is a **hue split** (the forehead one hue, the cheeks another, the jaw a third) caused by mixed lighting, makeup, or a heavy-handed adjustment. Sample five points with a 5×5 average eyedropper and compare.\n\n**4. Compare against something real.** A reference photograph of the same person under known light. A colour checker in the first frame. The person's own description. Nothing else is evidence.\n\n**5. Judge at the size it will be seen**, against a neutral grey background rather than a white one. A face against white looks darker and more saturated than the same face against grey.\n\n## The history, because it explains the tools\n\nThis matters for understanding why automatic tools behave as they do.\n\nColour film stocks and printing processes were calibrated for decades against reference cards (the \"Shirley cards\") showing a white woman. Laboratory printing equipment was set up so that card came out correctly, which meant the entire chain was optimised for light skin. Darker skin was reproduced poorly, with detail lost in the shadows, and this was a well-documented, widely-experienced failure rather than an occasional one.\n\nKodak made significant improvements to the shadow response of its colour stocks in the 1970s and 80s. The documented drivers included complaints from furniture and chocolate manufacturers, who could not get dark wood and dark chocolate to reproduce, as much as from photographers of people. Multiracial reference cards appeared later.\n\nThe pattern repeated in digital. Automatic white balance, automatic exposure and face detection are all built by measuring performance against datasets, and where those datasets have under-represented darker skin, the resulting systems have performed worse on it. This has been measured repeatedly in published audits of commercial face-analysis systems, with error rates differing substantially across skin tone and gender groups.\n\nThe practical consequence for you at the editing desk: **automatic exposure, automatic white balance, automatic \"portrait\" modes and one-click skin tools were tuned on a distribution, and if your subject sits outside it, the automatic answer will be systematically wrong in a consistent direction.** Knowing that turns a vague dissatisfaction into a specific correction.\n\n## Practical corrections that do not flatten\n\n**Exposure for dark skin.** The commonest error is underexposure, because the camera's meter reads a dark subject and pulls the exposure down. Expose for the skin, not the average: spot meter on the face, or add a stop and check the histogram. Lifting underexposed dark skin in post brings up noise and loses the specular highlights that give skin its life.\n\n**Do not chase a target lightness.** Lightening dark skin to match a reference of light skin is not a correction; it is an alteration of the person. The same applies to the automatic \"beautify\" modes in phone cameras that lighten and smooth by default, which many phones apply without asking.\n\n**Saturation, carefully.** Dark skin often carries more colour variation than a global saturation slider handles well, and pushing saturation globally shifts it towards red. A targeted Hue/Saturation on Reds and Yellows, or better, a perceptual chroma control like GIMP's Hue-Chroma or darktable's colour balance RGB, keeps lightness stable.\n\n**Specular highlights matter more, not less.** On darker skin the highlights carry much of the form. Crushing them with a heavy highlight recovery flattens the face. Keep them and control them locally with burning instead.\n\n**Mixed skin tones in one frame** (a family group, a team photograph) cannot all be optimised globally. Correct globally for neutral, then use masked local adjustments per person. This is normal professional practice and it is not a failure.\n\n## What to do when you are unsure\n\nAsk. If the photograph is of a person and you can reach them, show them two versions. This is not a cop-out; it is the only source of ground truth about how somebody's own skin looks, and it takes one message.\n\nFor commercial work, put a colour checker in the first frame of each setup. It converts an argument about taste into a measurement, and it costs about the price of a lens cap.\n\n## What remains unresolved\n\nThere is no agreed standard for \"correct\" skin reproduction across the full human range, and the industry references that exist were built for particular media and particular populations. Different display technologies, different printing processes and different viewing conditions all shift the answer.\n\nThe defensible position for an editor is procedural rather than numerical: neutralise from a known reference, keep hue consistent across the subject, preserve rather than flatten the tonal range, compare against a real reference where one exists, and be explicit about what you changed. That is a standard you can state and be held to, which is more than a set of CMYK ratios can offer."},"check":{"question":"An editor corrects a portrait of a dark-skinned subject by adjusting until the skin's CMYK values match a widely-quoted ratio derived from pre-press practice. The subject says it does not look like them. What is the most accurate assessment?","options":[{"text":"The ratio applies to print only, so the on-screen version will be correct once converted to CMYK for output.","correct":false,"tempting":"Treats a space conversion as the discrepancy when the numbers were already being matched in CMYK and the mismatch is with the person rather than with the medium."},{"text":"The subject is judging on an uncalibrated screen, so their impression is not reliable evidence about the correction.","correct":false,"tempting":"Dismisses the only ground truth available: display variation is real, and it does not license overriding a person's account of their own appearance."},{"text":"The ratio was derived from light skin on coated stock and does not describe the full range of human skin, so matching it has altered the subject rather than corrected the file.","correct":true},{"text":"The white balance must still be wrong, because a correctly neutralised file would naturally land on the standard ratio.","correct":false,"tempting":"Assumes the ratio is what neutral produces, when a correctly neutralised photograph of dark skin lands on quite different numbers by construction."}]},"prev":{"slug":"colour-grading-and-luts","title":"A look is a mapping, and a LUT is a table of it","url":"https://addaly.com/learn/image-editing/colour-grading-and-luts"},"next":{"slug":"matching-a-whole-set","title":"Forty photographs that have to look like one shoot","url":"https://addaly.com/learn/image-editing/matching-a-whole-set"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'There is no correct number for skin', in Editing an Image. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/image-editing/skin-tones-honestly (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"There is no correct number for skin","courseTitle":"Editing an Image","module":{"index":4,"slug":"tone-and-colour","title":"Tone and colour"},"answer":null,"keyIdea":"Skin has no universal correct value, so the method is to neutralise from a known reference, check hue consistency across the face rather than against a target, and recognise that automatic exposure and white balance were tuned on distributions that systematically under-serve darker skin.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"the-rule-you-will-be-told-and-its-actual-status","runs":[{"text":"The rule you will be told, and its actual status"}]},{"t":"p","runs":[{"text":"Search for skin tone correction and you will find a rule: in CMYK, yellow should be slightly higher than magenta, and cyan should be a fraction of magenta: often quoted as cyan being a quarter to a third of the magenta value, with magenta and yellow close together."}]},{"t":"p","runs":[{"text":"It is a genuine rule of thumb from the pre-press era, and it works within its range: it was derived from Caucasian skin printed on coated stock. Applied to darker skin, the ratios change substantially: cyan rises considerably relative to magenta, and the yellow-magenta relationship shifts. Applied as a universal target, it lightens and pinkens dark skin, which is exactly the failure mode this lesson is about."}]},{"t":"p","runs":[{"text":"The honest statement: "},{"text":"skin covers an enormous range of hues, lightnesses and saturations, and there is no single set of numbers that is correct.","b":true},{"text":" What there is instead is a method."}]},{"t":"h2","id":"the-method-that-works","runs":[{"text":"The method that works"}]},{"t":"p","runs":[{"text":"1. Get neutral first.","b":true},{"text":" Set white balance from something known to be neutral in the frame, as in the earlier lesson. Most “the skin looks wrong” problems are white balance problems, and fixing skin directly while the whole frame has a cast means fighting the same error twice."}]},{"t":"p","runs":[{"text":"2. Check that the cast is gone from something that is not skin.","b":true},{"text":" A white shirt, a grey wall, the whites of the eyes. If those are neutral and the skin still looks wrong, the problem is in the skin."}]},{"t":"p","runs":[{"text":"3. Look for hue consistency across the face rather than a target value.","b":true},{"text":" Skin should be one hue family across the whole face, varying in lightness and saturation. The commonest genuine fault is a "},{"text":"hue split","b":true},{"text":" (the forehead one hue, the cheeks another, the jaw a third) caused by mixed lighting, makeup, or a heavy-handed adjustment. Sample five points with a 5×5 average eyedropper and compare."}]},{"t":"p","runs":[{"text":"4. Compare against something real.","b":true},{"text":" A reference photograph of the same person under known light. A colour checker in the first frame. The person’s own description. Nothing else is evidence."}]},{"t":"p","runs":[{"text":"5. Judge at the size it will be seen","b":true},{"text":", against a neutral grey background rather than a white one. A face against white looks darker and more saturated than the same face against grey."}]},{"t":"h2","id":"the-history-because-it-explains-the-tools","runs":[{"text":"The history, because it explains the tools"}]},{"t":"p","runs":[{"text":"This matters for understanding why automatic tools behave as they do."}]},{"t":"p","runs":[{"text":"Colour film stocks and printing processes were calibrated for decades against reference cards (the “Shirley cards”) showing a white woman. Laboratory printing equipment was set up so that card came out correctly, which meant the entire chain was optimised for light skin. Darker skin was reproduced poorly, with detail lost in the shadows, and this was a well-documented, widely-experienced failure rather than an occasional one."}]},{"t":"p","runs":[{"text":"Kodak made significant improvements to the shadow response of its colour stocks in the 1970s and 80s. The documented drivers included complaints from furniture and chocolate manufacturers, who could not get dark wood and dark chocolate to reproduce, as much as from photographers of people. Multiracial reference cards appeared later."}]},{"t":"p","runs":[{"text":"The pattern repeated in digital. Automatic white balance, automatic exposure and face detection are all built by measuring performance against datasets, and where those datasets have under-represented darker skin, the resulting systems have performed worse on it. This has been measured repeatedly in published audits of commercial face-analysis systems, with error rates differing substantially across skin tone and gender groups."}]},{"t":"p","runs":[{"text":"The practical consequence for you at the editing desk: "},{"text":"automatic exposure, automatic white balance, automatic “portrait” modes and one-click skin tools were tuned on a distribution, and if your subject sits outside it, the automatic answer will be systematically wrong in a consistent direction.","b":true},{"text":" Knowing that turns a vague dissatisfaction into a specific correction."}]},{"t":"h2","id":"practical-corrections-that-do-not-flatten","runs":[{"text":"Practical corrections that do not flatten"}]},{"t":"p","runs":[{"text":"Exposure for dark skin.","b":true},{"text":" The commonest error is underexposure, because the camera’s meter reads a dark subject and pulls the exposure down. Expose for the skin, not the average: spot meter on the face, or add a stop and check the histogram. Lifting underexposed dark skin in post brings up noise and loses the specular highlights that give skin its life."}]},{"t":"p","runs":[{"text":"Do not chase a target lightness.","b":true},{"text":" Lightening dark skin to match a reference of light skin is not a correction; it is an alteration of the person. The same applies to the automatic “beautify” modes in phone cameras that lighten and smooth by default, which many phones apply without asking."}]},{"t":"p","runs":[{"text":"Saturation, carefully.","b":true},{"text":" Dark skin often carries more colour variation than a global saturation slider handles well, and pushing saturation globally shifts it towards red. A targeted Hue/Saturation on Reds and Yellows, or better, a perceptual chroma control like GIMP’s Hue-Chroma or darktable’s colour balance RGB, keeps lightness stable."}]},{"t":"p","runs":[{"text":"Specular highlights matter more, not less.","b":true},{"text":" On darker skin the highlights carry much of the form. Crushing them with a heavy highlight recovery flattens the face. Keep them and control them locally with burning instead."}]},{"t":"p","runs":[{"text":"Mixed skin tones in one frame","b":true},{"text":" (a family group, a team photograph) cannot all be optimised globally. Correct globally for neutral, then use masked local adjustments per person. This is normal professional practice and it is not a failure."}]},{"t":"h2","id":"what-to-do-when-you-are-unsure","runs":[{"text":"What to do when you are unsure"}]},{"t":"p","runs":[{"text":"Ask. If the photograph is of a person and you can reach them, show them two versions. This is not a cop-out; it is the only source of ground truth about how somebody’s own skin looks, and it takes one message."}]},{"t":"p","runs":[{"text":"For commercial work, put a colour checker in the first frame of each setup. It converts an argument about taste into a measurement, and it costs about the price of a lens cap."}]},{"t":"h2","id":"what-remains-unresolved","runs":[{"text":"What remains unresolved"}]},{"t":"p","runs":[{"text":"There is no agreed standard for “correct” skin reproduction across the full human range, and the industry references that exist were built for particular media and particular populations. Different display technologies, different printing processes and different viewing conditions all shift the answer."}]},{"t":"p","runs":[{"text":"The defensible position for an editor is procedural rather than numerical: neutralise from a known reference, keep hue consistent across the subject, preserve rather than flatten the tonal range, compare against a real reference where one exists, and be explicit about what you changed. That is a standard you can state and be held to, which is more than a set of CMYK ratios can offer."}]}],"check":{"question":[{"text":"An editor corrects a portrait of a dark-skinned subject by adjusting until the skin’s CMYK values match a widely-quoted ratio derived from pre-press practice. The subject says it does not look like them. What is the most accurate assessment?"}],"options":[{"text":[{"text":"The ratio applies to print only, so the on-screen version will be correct once converted to CMYK for output."}],"correct":false,"tempting":[{"text":"Treats a space conversion as the discrepancy when the numbers were already being matched in CMYK and the mismatch is with the person rather than with the medium."}]},{"text":[{"text":"The subject is judging on an uncalibrated screen, so their impression is not reliable evidence about the correction."}],"correct":false,"tempting":[{"text":"Dismisses the only ground truth available: display variation is real, and it does not license overriding a person’s account of their own appearance."}]},{"text":[{"text":"The ratio was derived from light skin on coated stock and does not describe the full range of human skin, so matching it has altered the subject rather than corrected the file."}],"correct":true,"tempting":null},{"text":[{"text":"The white balance must still be wrong, because a correctly neutralised file would naturally land on the standard ratio."}],"correct":false,"tempting":[{"text":"Assumes the ratio is what neutral produces, when a correctly neutralised photograph of dark skin lands on quite different numbers by construction."}]}],"order":[0,2,3,1],"why":null},"prev":{"slug":"colour-grading-and-luts","title":"A look is a mapping, and a LUT is a table of it","minutes":9},"next":{"slug":"matching-a-whole-set","title":"Forty photographs that have to look like one shoot","minutes":8}}}