Notes from a quiet darkroom
AI photo editing, explained slowly
What an AI editor actually changes in a picture, how removal, relighting and retouching work underneath, and how to keep an edited photo honest about what it was.
On the contact sheet
What the tools change in a picture
Most AI edits fall into a handful of kinds. Each one alters something different, and each asks a different question about whether the result still tells the truth.
-
Removal
Taking out a passer-by, a power line or a stray bin. The tool does not uncover what was behind it; it paints a plausible guess into the gap.
-
Relighting
Moving or softening the light after the shutter has closed. Shadows, highlights and the colour of the light all have to agree for it to hold up.
-
Retouching
Dust, blemishes, flyaway hair and creases. Small repairs of the kind a darkroom printer once made with a fine brush, now made in a moment.
-
Backgrounds
Blurring, cleaning or replacing whatever stands behind the subject. The hard part is the edge: hair, glass and anything half see-through.
-
Expanding the frame
Growing a picture past its own borders to fit a wider or taller shape. Everything outside the original frame is invented from what was inside it.
-
Colour and tone
White balance, contrast and grading. The oldest kind of edit, and still the one that changes the mood of a picture most with the least fuss.
Under the safelight
How an AI edit is made
Behind a single button, most edits pass through the same three stages. Knowing them makes it easier to see where a result has gone wrong.
- Selection
- The tool first decides which pixels belong to what: the person, the sky, the object you tapped. A loose selection is where halos and missing fingertips begin.
- Generation
- New pixels are predicted from the ones around them. They are a guess at what could be there, never a recovery of what was there.
- Blending
- The new area is matched to the old in grain, sharpness and light. When blending falls short, the patch shows as a smooth or oddly lit island.
Filed with the negatives
Keeping an edited photo honest
A changed picture is not a dishonest one. It becomes one when it is passed off as a record of something that did not happen.
-
Keep the original
Save the untouched file beside the edit. It settles questions later, and it lets you start again when an edit goes too far.
-
Say when it was changed
Where a picture is offered as fact, a short note that it was edited, and how, keeps the reader's trust in everything else you show them.
-
Leave the evidence alone
News, insurance claims, property listings and anything used as proof call for pictures that show what was really there.
-
Read the rules first
Competitions, publishers, marketplaces and social platforms often set their own terms for generated or heavily edited images. They are worth checking before you submit.
Articles on these subjects are on the way.
What readers ask
The questions it is organised around
- Are there AI photo editor online free tools that work without a signup?
- Can you recommend a free AI photo editor that doesn't watermark exports?
- What's the fastest way to AI edit photo backgrounds across a product catalog?
- Which AI photo editor free tier still allows high resolution downloads?
- Is there an AI picture editor simple enough for someone who's never edited before?
- Which image editor AI tools integrate with a RAW workflow?
- Which AI photo editor handles retouching without making skin look plastic?
- How much can photo editor AI features replace manual masking work?
- What AI image editor is best for batch correcting a whole shoot?
- What free AI image editor works well on a low spec laptop?