There is a version of AI editing that photographers are right to distrust: the one that takes a gallery, applies a look nobody chose, and hands back images that are technically fine and unmistakably not yours.
And there is a version that is simply a very fast assistant who has watched you work for years, does the first eighty percent, and leaves the decisions to you.
The difference is not the model. It is how much control survives the process. This guide covers what AI photo enhancement actually does, where the sensible line sits between automatic and manual, and how to get speed without handing over your style.
What AI photo enhancement actually does
Strip away the marketing and there are three separate jobs being sold under one name.
Correction. Exposure, white balance, contrast, recovery of highlights and shadows. Broadly objective work — there is usually a defensible right answer, and this is where automation is least controversial and most useful.
Interpretation. Applying a look. Colour grading, tonal character, the warmth or coolness that makes a gallery feel like it came from you. Not objective at all, and where AI most often goes wrong, because it is guessing at taste.
Retouching. Skin, blemishes, stray hairs, distractions. Subjective in degree rather than direction — everyone agrees a spot goes, nobody agrees how much skin texture should survive.
Most tools do all three and describe it as one feature. Judging them separately is the fastest way to work out whether a tool suits you, because a photographer who wants automated correction and manual interpretation is very well served, and one who wants all three automated is buying a different product.
Where the line sits between automatic and manual
The honest answer, and the one most software will not give you: automate the corrections, keep the interpretation.
Corrections are repetitive, measurable and boring. There is no craft in equalising exposure across forty frames shot as the sun went behind a cloud. That is arithmetic, and handing it over costs you nothing.
Interpretation is the thing clients hire you for. It is also the thing an AI cannot infer from the file alone, because the information it needs is not in the pixels — it is in eight years of your work and a conversation you had with the couple in March.
In practice this means the useful pattern is not "AI edits, you approve". It is AI gets you to a defensible starting point, you spend your time on the frames that matter. ShootCleaner's enhancement stage is built as style-aware starting points with manual control retained, which is the same idea stated as a product decision: finish most of it automatically, fine-tune the rest.
The number people quote is around eighty percent, and it is a reasonable rule of thumb. The last twenty percent is where the photographs are.
Style: the part that actually decides whether you keep using it
Every AI editing tool in this market answers the same objection unprompted somewhere in its marketing: will it still look like my work? When every competitor pre-empts the same question, that tells you it is the fear that blocks the sale.
There are three ways tools try to answer it, and they are not equally good.
Generic presets. A library of looks you pick from. Fast, and the reason so much AI-edited work looks like other AI-edited work. Fine as a starting point, useless as an identity.
Training on your catalogue. The tool learns from thousands of your past edits. Genuinely effective, and slow to set up — you need a large, consistent back catalogue, and if your style has changed in the last two years the model learns the average of two styles rather than your current one.
Reading your portfolio. ShootCleaner builds a Signature Look profile from your portfolio — the work you have chosen to show, which is a much sharper signal than everything you have ever exported. Your portfolio is the edit you already made about what represents you.
That last distinction matters more than it sounds. Your catalogue contains every job, including the ones shot to a client's brief you disagreed with. Your portfolio contains the work you want more of. One is a record, the other is an intention.
Presets still matter
None of this replaces presets, and any tool that tells you it does is overselling.
A preset is a deterministic, portable, inspectable set of adjustments. It does the same thing every time, it moves between machines, and you can see exactly what it changed. That reliability is worth keeping.
What presets cannot do is adapt. The same preset applied across a shoot where the light shifted will be right in the middle and wrong at both ends, which is why "apply preset, then fix every frame by hand" is the workflow most photographers actually run.
The productive combination is both: the preset carries your look, the AI handles the per-frame variance the preset cannot see. ShootCleaner works with your existing Adobe XMP presets rather than asking you to rebuild them, so the look you have spent years refining is the input rather than something to recreate.
Non-destructive editing, and why it is not optional
Non-destructive editing means your original file is never modified. Edits are stored as instructions — a list of adjustments applied on the way out — rather than baked into pixels.
The practical consequences:
- Nothing is ever lost. Reset to original at any point, however many rounds in.
- You can change your mind cheaply. A client asking for a warmer set six months later is a re-export, not a re-edit.
- Versions cost nothing. Colour and black-and-white are two instruction sets over one file, not two files.
- Mistakes are not permanent. A bad batch is undone, not repaired.
This becomes more important the more automation you add, not less. If you are letting software make the first pass on three thousand frames, the ability to reverse that pass entirely is what makes it safe to try. Destructive automated editing is a category that should not exist.
The question worth asking any tool: when it "enhances" an image, is it writing a new file or writing instructions? If the answer involves exporting a JPEG and re-importing it, you have left non-destructive territory — and you have also just re-encoded every photograph in the shoot.
Consistency across a shoot, and across months
Two different problems, and the second is harder.
Within a shoot, consistency means the gallery reads as one event. The reception frames and the ceremony frames should feel related even though the light was completely different. This is where per-frame AI correction earns its place: it is normalising toward a common target rather than applying an identical adjustment to dissimilar frames.
Across months, consistency means a client who booked you in March and again in November gets the same photographer. This is the one studios underrate, and it is a business problem rather than an aesthetic one — a returning client noticing that this year's photographs look different from last year's is a retention risk.
It is also acute for multi-photographer studios. Two photographers, fifteen sessions a week, each culling and editing to slightly different instincts, produces a studio with no house style. A shared profile applied across both is worth more than either photographer's individual speed.
Retouching without the plastic look
The failure mode of automated retouching is instantly recognisable: skin like moulded plastic, every pore gone, faces that look printed rather than photographed.
It happens because aggressive smoothing is the easiest thing to automate and the easiest to over-apply, and because default settings tend to be tuned for a dramatic before-and-after rather than a good photograph.
What to look for:
- Texture retained. Skin should still read as skin at 100%. If pores are gone, it is too much.
- Strength control that is actually granular. A slider, not a toggle.
- Frequency-aware work. Good retouching removes blemishes without flattening the underlying structure of a face.
- Per-subject sanity. A newborn, a bride and a seventy-year-old grandfather should not receive identical treatment.
The reliable test: process a frame, then look at it beside the original at full magnification on a decent screen. If your eye goes straight to the skin, the retouching is doing too much. Good retouching is invisible.
Desktop or cloud?
A genuine fork, and it comes down to three things.
Speed at volume. Cloud tools upload your images before they process them. On a five-thousand-frame event over a domestic connection, the upload is frequently longer than the editing. Desktop tools start immediately, and the ceiling is your own hardware rather than a shared queue.
Predictability. Local processing takes the same time on Tuesday morning as on Sunday night. Cloud services vary with load, and Sunday night is exactly when photographers are working.
Privacy. Client photographs leaving your machine is a question worth having a defensible answer to, particularly for weddings, newborns and anything involving children. For some contracts it is a requirement rather than a preference.
The counter-argument is real: cloud tools need no hardware and work from anywhere. If you edit on a laptop in cafés, that matters. If you edit a 3,000-frame wedding on a desktop at home, it does not.
ShootCleaner processes locally on Windows 10 and 11, with a fourteen-day offline grace period, so a week on location with no connection does not stop you working. macOS is in early access.
What the AI costs you, and who pays for it
Easy to overlook, and it changes the arithmetic at volume.
Most AI editing tools bundle the model cost into a subscription. That is convenient, and it means you are paying the vendor's margin on top of the underlying AI cost, usually metered as credits or per-image charges you cannot audit.
The alternative is bringing your own key. ShootCleaner uses your own OpenAI or Google Gemini key, encrypted in your system keychain and never stored on a server. You pay the provider directly at their rates, with no markup and no per-image credit meter. It is marginally more setup and materially cheaper at volume — and it means your AI costs are visible, on a bill you control.
The licence itself is €195 one-time for two devices, with twelve months of Connected Care included and optional renewal at €95 a year. If you do not renew, the version you own keeps working. There is a fourteen-day refund window.
A workable order of operations
- Cull first. Never edit before you have selected. Editing frames you will not deliver is the most common waste in post-production.
- Correct globally. Let automation normalise exposure and white balance across the set.
- Apply your look. Signature Look profile, your existing XMP presets, or both.
- Review at speed. Scan the set for outliers. You are looking for frames the automation misread, not perfecting anything yet.
- Fine-tune the frames that matter. The cover image, the first dance, the one the client will print. These deserve manual time, and now you have it.
- Retouch selectively. Not the whole gallery. The frames where a person will look at their own face.
- Export once. Every extra export is another re-encode of the same photograph.
Step one and step seven are where most of the recoverable time actually is, and neither has anything to do with editing.
Frequently asked questions
Will AI editing make my work look generic? Only if the tool has no information about your style. A profile built from your own portfolio, or your own presets, is what prevents it. Ask any tool where its sense of your look comes from.
Can I still edit manually? You should be able to, on every frame. Automation that cannot be overridden is not a workflow, it is a filter. ShootCleaner's enhancement stage retains manual fine control throughout.
Does AI editing replace Lightroom? Not for raw development or serious retouching, and be sceptical of anything claiming otherwise. It removes the repetitive first pass. Most photographers keep Lightroom and many are happy to.
Is it non-destructive? It must be. Your originals should never be modified, and any automated pass should be fully reversible.
Can I use my existing presets? With ShootCleaner, yes — it works with Adobe XMP presets you already own rather than requiring you to rebuild your look.
What does it cost to run? With a bring-your-own-key model, you pay your AI provider directly at their rates with no markup. The software itself is a one-time €195 licence rather than a subscription.
The part editing does not solve
Worth ending here, because it is the thing that actually eats evenings.
You can cut editing time by seventy percent and still finish at midnight, because editing is one stage. Around it sit the export, the upload, the proof gallery, the wait for the client, the reconciliation of their filename list back into a selection, the second edit, the second export and the second upload.
Most photographers run that across six or seven applications that were never built to talk to each other. Optimising the editing step while leaving the handoffs alone is optimising the part that was already the most enjoyable.
Culling is one decision. Delivery is the whole job.
ShootCleaner runs cull, enhance, metadata, presentation and delivery as a single pass on your own machine — no exports in the middle. €195 once, two devices. Explore the workflow or download for Windows.