AI image editing has stopped being a novelty and become a set of ordinary jobs: removing a light stand from a product photo, filling in the edge of a cropped frame, enlarging a small image for print, cutting a background out cleanly, or editing a whole catalogue by instruction rather than by hand. The tools that do this range from free browser editors to subscription suites whose generative features are metered by credits. This guide sorts the jobs from the marketing, compares what a free tier really allows against what a paid plan adds, and gives a one-hour test that tells you whether a tool is worth paying for.
Almost every product on the market does some subset of six tasks. Identify which of them you actually repeat before comparing anything else.
The licence and the limits matter more than the feature list. These are the differences that show up across most tools.
| What a free tier typically gives | What a paid plan usually adds | What to check before you rely on it |
|---|---|---|
| Watermark on export, or editor-only preview | Clean exports with no watermark | Whether the watermark appears on preview only or on the saved file |
| Resolution capped, often around a few megapixels | Full-resolution export suitable for print | The exact pixel dimensions your print or platform needs |
| A small monthly credit allowance for generative edits | A larger allowance, or metered pay-as-you-go | How many credits one edit consumes, and what happens at the cap |
| Personal or evaluation licence | Commercial licence and, in some cases, indemnity | Whether your use is commercial, and what the terms say about it |
| One image at a time, manual upload | Batch queues, API access, shared libraries | Whether your workflow needs volume or repeatability |
None of this is hidden, but it is rarely on the pricing page. Read the terms for commercial rights and export limits before you promise a client anything.
Free does not mean amateur. For crop, clean-up, cut-outs and simple retouching, the free layer is genuinely enough for a lot of work.
Features, limits and licences in all four categories change often, so verify the current terms on the tool page rather than relying on a review, including this one.
Buy the smallest plan that covers the job you repeat. A single-purpose tool that does one task well is often cheaper and faster than a suite whose three other modules you never open.
Uploaded images are the real privacy question. A cloud tool receives your file, and the terms decide whether it is stored, for how long, and whether it can be used to train models. For confidential material - client products, unreleased designs, images of people - prefer a desktop tool that processes locally, or a business plan with a data-processing agreement.
Disclosure is the second question, and it is becoming a compliance topic rather than a courtesy. Many platforms label synthetic or significantly manipulated media, and advertising standards in several markets treat misleading edits differently from cosmetic ones. Removing a stray object is not the same as adding a person who was never in the photograph. Content-provenance standards such as C2PA aim to record editing history in the file metadata, and some publishers now require it.
The source material matters too. Editing an image does not give you rights to it, and generating a likeness of a real person raises questions that a licence does not settle. Keep the original files, note what was changed, and when in doubt ask before publishing.
It depends on the task and the licence. A free browser or open-source editor can handle cropping, clean-up, background removal and simple retouching to a professional standard. What free tiers usually restrict is commercial usage rights, maximum export resolution, batch processing and generative features that consume credits. Read the terms for the specific tool before you deliver client work made with it.
Typically four things: higher export resolution, a commercial licence, batch or API access, and generative features with a monthly credit allowance instead of a watermark. Paid plans also tend to add non-destructive layers, version history and shared team libraries. If you need only one of those, check whether a cheaper single-purpose tool covers it before buying a full suite.
Only if the terms allow it. Check where files are stored, whether they are used to train models, how long they are retained, and who can see them. Some tools process images locally on your device, which avoids the question entirely. For confidential material, prefer a desktop tool or a business plan with a data-processing agreement.
It depends on where the image appears and what changed. Many platforms require labels for synthetic or significantly manipulated media, and some jurisdictions regulate misleading edits in advertising. Brightening an image or removing a stray object is not the same as adding a person who was never there. When in doubt, check the platform rules and the advertising standards that apply to you.
Good tools work non-destructively: the edit lives in a layer or an instruction stack that you can switch off, and the original file is untouched. Generative edits that replace pixels are harder to reverse, so keep an unedited copy of the source, work on a duplicate, and note what was changed.
References - official project and tool documentation (checked September 2026)
Last updated: 2026-09-13. Free-tier limits, licences and AI features change frequently; verify current details on the official pages above before you rely on a tool for paid work.