Lifting shadows brings up noise and colour shift together, and skin shows both first. The order you do things in decides whether the recovery is usable.
The short answer: lift conservatively from the raw, fix the shadow colour shift before anything else, denoise before you sharpen anything, then rebuild contrast locally rather than globally. Skin gets retouched after the recovery, not before. Doing those in the wrong order is what makes a recovered frame look recovered.
Underexposed portraits are recoverable far more often than photographers expect, and ruined in recovery more often than they should be. The difference is almost entirely the order of operations.
What actually happens when you lift shadows
Three things at once, and skin shows all three first.
Noise appears. It was always there; you are amplifying it along with the signal. In shadows the ratio of signal to noise is worst, which is exactly where you are lifting.
Colour shifts. Sensor response is not perfectly linear at the bottom of the range. Deep shadows commonly go green or magenta as they come up, and on a face that reads immediately as wrong.
Contrast collapses. Lifting shadows without touching anything else compresses the tonal range. The result is flat and grey, and the instinctive fix — adding contrast — pushes the noise straight back up.
The order that works
1. Start from the raw
Non-negotiable. A JPEG has already discarded most of the shadow information and has compression artefacts in exactly the region you are about to amplify.
2. Lift globally, but conservatively
Exposure and shadows in the raw processor, before anything else. Do not try to reach the final brightness here — get to about 80% of it. The last 20% is done locally, where you can protect what needs protecting.
3. Fix the colour shift immediately
Before denoising, before contrast. Look at the shadow areas: if they have gone green or magenta, correct that now.
Doing it later means every subsequent decision was made against a wrong colour, and you will end up correcting twice.
4. Denoise before anything sharpens
Machine-learning denoise is genuinely transformative here — it is the tool that turned a category of frames from unusable into deliverable.
Two rules: run it at full resolution, and check skin specifically. Denoise removes fine detail, and skin texture is fine detail. Over-denoised skin is plastic skin by a different route.
5. Rebuild contrast locally, not globally
Global contrast raises noise everywhere including the areas that were fine. Instead: a gentle S-curve on the mid-tones, then local dodge and burn to put the modelling back where the lift flattened it.
6. Retouch the skin normally
Now, and not before. Frequency separation, dodge and burn, the usual process. Doing skin work on an unrecovered file means redoing it after the recovery.
7. Sharpen selectively, at the end
Masked to eyes and edges. Never globally — you will sharpen the noise you just spent effort removing.
How far you can push it
Roughly, on a modern full-frame sensor with a raw file:
One stop — recovers cleanly, almost invisibly. Two stops — recovers well with denoise. This is normal territory. Three stops — recoverable for web delivery, harder for print. Expect visible work. Four or more — you can produce something, but colour in the shadows is unreliable and detail is largely gone.
Crop sensors and older cameras have less room. Test your own body so you know your number rather than someone else's.
The specific problem of dark skin
Deeper skin tones sit lower in the tonal range, so they are more affected by every one of the issues above, and the colour shift matters more because there is less separation between the skin and the shadow.
Two things help: correct the colour shift before doing anything else — it is the dominant problem, not the brightness — and use luminosity masks that target the skin's actual range rather than a broad "shadows" selection that also catches the background.
Never lift dark skin toward a lighter tone in the name of "brightening". The subject's tone is the subject's tone. Brighten the exposure, not the person.
When to accept it as a dark frame
Some underexposed portraits are better as low-key photographs than as recovered ones. If the face is lit and the rest is dark, that can be the picture — grade it deliberately, keep the shadows shadowed, and it reads as a choice.
Recovery for its own sake produces a grey frame with noise in it. A deliberate low-key frame is a photograph.
Preventing it
Expose for the face, not the average. Use highlight-weighted metering if your camera has it. When in doubt on a raw file, exposing slightly to the right and pulling back in post is nearly always better than the reverse.
What we do with an underexposed set
Recovery is part of the ordinary work at every tier — it is not an add-on. What matters is that the shadow colour is corrected before the skin is touched and that denoise runs before any sharpening, because doing those in the wrong order is what makes recovered frames look recovered.
If a frame is beyond honest recovery we say so rather than delivering a noisy grey version of it. That is worth more to a photographer than a false yes.
Knowing when to stop
The most useful thing to develop here is a sense of your own camera's limit, because it is a specific number rather than a general one.
Take a frame you do not need, expose it correctly, then shoot it at one, two, three and four stops under. Recover each and compare. You will find the point where the colour in the shadows stops being reliable, and after that you will never have to guess on a job.
Below the limit, recovery is ordinary work and nobody will know. Above it, you are producing something rather than recovering something, and the honest move is to say so.
Some frames are better as deliberate low-key photographs than as recovered ones. If the face is lit and the rest is dark, grade it that way — keep the shadows shadowed, and it reads as a choice instead of an accident.
Common questions
How many stops can I recover? Roughly: one stop cleanly, two well with denoise, three for web delivery with visible work, four or more producing something rather than recovering something. Test your own camera so you know your number rather than someone else's.
Why does my recovered image look green or magenta in the shadows? Sensor response is not perfectly linear at the bottom of the range. Correct that shift before doing anything else — every later decision made against a wrong colour has to be made twice.
Should I denoise before or after lifting? Lift first, conservatively, then correct colour, then denoise. Denoising before the lift removes detail you have not yet seen; denoising after sharpening amplifies what you were removing.
Is recovering dark skin different? It needs more care rather than a different method. The colour shift matters more because there is less separation from the shadow, so correct hue before brightness — and never lift a subject's skin toward a lighter tone in the name of brightening.
When should I leave it dark? When the face is lit and the rest is not. Graded deliberately, that is a low-key photograph. Recovered for its own sake it is a grey frame with noise in it.
Related reading: How to Fix Mixed Lighting and Wrong Skin Tones and RAW vs JPEG for Professional Retouching.
BeEdits Editorial Team
Post-production studio
A photography post-production company: high-end retouching, creative editing and scalable production support for photographers, studios and brands.
