Photo editing and retouching

Why RAW files give a genuinely better result than JPEGs

A RAW file is everything the camera sensor recorded. A JPEG is the camera guessing what you wanted and throwing the rest away. That discarded information is exactly what retouching needs.

The practical difference shows up in three places. Exposure recovery: a RAW file typically holds several stops of detail in highlights and shadows that a JPEG has already flattened, so a slightly overexposed sky can be brought back rather than being a white shape. White balance: RAW stores colour temperature as an adjustable setting, so a photo shot under tungsten light can be corrected perfectly after the fact, while a JPEG has the wrong colour baked in and can only be pushed so far before skin goes grey. Bit depth: RAW commonly carries 12 or 14 bits per channel against 8 in a JPEG, which means far more tonal steps and no banding when a gradient or a sky gets adjusted.

None of this means JPEGs are unusable. Plenty of good work starts from them, especially phone images. It means the ceiling is lower and heavy corrections start to show as posterised gradients and muddy colour. If a shoot is being commissioned, ask for RAW plus JPEG, and keep the RAW files even after delivery.

How professional skin retouching actually works

Good skin retouching removes distractions and leaves the person recognisable. The failure mode everyone knows, the plastic, poreless look, comes from blurring skin, which destroys texture along with the blemish.

The professional approach separates temporary from permanent. Temporary things go: a spot that will be gone next week, a stray hair across the face, a lipstick smudge, a shine from studio lights. Permanent things stay unless the subject has specifically asked otherwise: freckles, moles, scars, laugh lines, the shape of a nose. That distinction is both an ethical position and a practical one, because removing permanent features is what makes an image stop looking like the person.

Technically, the work is done at two levels rather than one. Blemishes are removed at texture level with healing tools that sample nearby skin, preserving pores. Uneven tone and blotchiness are smoothed at colour level separately, which evens out redness without touching detail. This separation, often called frequency separation, is why professional retouching holds up at full resolution while a beauty filter falls apart the moment you zoom in. Dodge and burn is then used sparingly to even out shadows that make skin look bumpy, working on light rather than on the skin itself.

• Remove what is temporary, keep what makes the person themselves
• Work on texture and colour separately, never blur skin wholesale
• Dodge and burn to even shadows rather than paint over them
• Check at 100 percent zoom and at final display size, both matter
• Agree in advance how far retouching should go, and put it in writing

Cutouts, masking and the difference hair makes

Removing a background sounds like one job but is really three, and which one applies decides how long it takes and what it costs.

Hard-edged subjects, meaning products, furniture, packaging, anything with clean boundaries, are cut with paths. The result is a crisp, reusable edge that scales cleanly and can be adjusted later. Soft-edged subjects, like fabric, fur, foliage or motion blur, need masking with edge refinement so the transition stays natural rather than looking stamped on. Hair is the hardest case in the field, because individual strands are semi-transparent and pick up colour from whatever was behind them.

That last point is why a cutout can look wrong even when the mask is technically accurate. A person photographed against a green wall has faint green in the edge of their hair. Place them on white and that fringe reads as a dirty halo. Proper work includes decontaminating those edge colours and often rebuilding a few strands so the silhouette does not look shaved. If cutouts are a regular need, photographing against a neutral grey rather than a saturated backdrop makes every later cutout easier.

Keeping hundreds of images consistent

Consistency across a large set is a process problem, not a talent problem. One image edited beautifully is straightforward. Four hundred images that look like they belong together requires a system.

We start by editing a small reference set, usually five to ten images that cover the range of the shoot: brightest, darkest, warmest, coolest, and a couple of typical ones. Those get approved first and become the standard. From them we build a preset or profile that carries the base corrections, then apply it across the batch as a starting point rather than a finished result. Every image still gets individual attention, because a preset cannot know that one photo was shot in shade.

The check at the end matters as much as the edit. Images are reviewed as a contact sheet, all together at small size, which is how inconsistency becomes obvious. A single image that is half a stop brighter is invisible on its own and jumps out immediately in a grid. We also keep the written style note: target brightness, contrast approach, saturation level, skin tone direction and sharpening amount. That note is what lets a second editor join the project six months later and match the first.

• Approve a reference set covering the range before touching the batch
• Use presets as a starting point, never as the finished edit
• Review the full set as a contact sheet, not one image at a time
• Keep a written style note so future work matches
• Spot check at full resolution on a calibrated screen

Colour spaces, file formats and exporting for web and print

The same edited image needs different files for different destinations, and sending the wrong one is the usual cause of colours that look dull online or wrong in print.

Colour space is the first decision. sRGB is the safe standard for anything on the web, because browsers and social platforms assume it and an image tagged in a wider space often displays flat and desaturated. Adobe RGB and ProPhoto RGB hold more colours and are useful while editing and for print work, but should be converted to sRGB on export for screen. Print work goes to CMYK only at the end, and ideally to the profile the specific printer asks for, because the conversion is not reversible and different presses differ.

Format follows use. JPEG for photographs on the web, at a quality setting high enough to avoid visible artefacts around edges. PNG only where transparency is needed or for flat graphics, because it is much heavier for photographs. WebP and AVIF give considerably smaller files at the same visual quality and are well supported now, which matters for page speed. TIFF or PSD as the archive and print master, keeping layers so the edit can be revisited. We deliver whichever set you need, and we keep a layered master regardless, because flattened files cannot be adjusted later without starting over.

• sRGB for anything on screen, always
• Convert to CMYK only at the final print step, using the printer profile
• JPEG for web photos, WebP or AVIF where page speed matters
• PNG only for transparency or flat graphics
• Keep a layered TIFF or PSD master for every image

Where AI tools help and where they still fail

Automated editing tools have become genuinely useful for parts of this work, and pretending otherwise wastes client money. Being honest about their limits matters just as much.

They are strong at repetitive, well-defined tasks: initial subject selection, straightforward background removal on clean shots, upscaling a small image, removing sensor dust, and filling a small area of plain background when an image needs extending to a different crop. Used as a first pass on a large batch, they save real time and the saving should reach the client rather than being quietly absorbed.

They are unreliable where judgement is involved. Automatic skin smoothing routinely erases texture and character. Generated fill invents detail that was never there, which is a problem the moment an image represents a real product or a real place. Edge detection struggles with hair, glass, reflections and anything semi-transparent. And automatic colour tools have no idea what your brand should look like. Our position is simple: automation handles the mechanical steps, a person makes every decision that affects how something looks or whether it is truthful, and nothing generated gets added to an image that represents a real product without you knowing.