The aggressive pass, and the right first choice when all you care about is whether the file gets through. On difficult sources it can smear fine detail or shift color, which shows up first on small or distant faces.
Forensic Cleaner by Two Tensors
Remove AI detection from your image
What you need before you start, which settings actually change the result, and what to do when a detector still flags the file.
Most images need one upload and one decision.
Open the cleaner, drop the file in, wait for validation, run it, download the result. The rest of this page exists for the cases where that is not enough.
Sign in and open Forensic Cleaner, then put your file on the workspace. Validation runs before anything is spent: it checks the dimensions, applies the safety checks, and returns an AI slop score, which is an estimate of how hard this particular image will be to clean.
Then pick a strength and run it. Strong is the default because it is the setting most likely to work on the first attempt. When the job finishes you get the cleaned copy next to the original, and the run stays in your account so you can pull it down again, rerun it at a different setting, or remove it entirely.
That is the loop. If you want the same ground covered one control at a time, the step by step guide is slower and more literal. If you want to know what the detectors are actually measuring before you touch anything, how image detector bypass works is the better starting point.
Before you start
- An image you made, or one you hold a license to process.
- A PNG, JPG, or WEBP file. Convert anything else first.
- An account. The free tier includes 10 credits a month.
- A detector you care about, so you can test the export instead of guessing.
The four controls, and when to move off the default.
The gentler pass, on paid plans. Use it when Strong worked but the output no longer looks like your photograph. It touches the image less, which also means it leaves more behind, so it is a second attempt rather than a default.
Free accounts export at up to 768px on the long edge and paid plans at up to 1024px, with the pipeline topping out near 2 megapixels. A smaller export is not automatically a worse result.
Watermark removal is a separate optional stage with its own strength setting, not part of the main clean. Medium is a reasonable place to begin, and lower settings distort small faces less.
Reading the numbers, before the run and after it.
The slop score is a difficulty estimate, not a verdict
Validation hands back an AI slop score before you commit anything. It is the app's estimate of how strongly this image reads as generated, and therefore how hard it will be to clean. A high score is not a prediction that the run will fail. It tells you to start on Strong, to expect the cleaning effect to be more visible in the output, and to plan for a second attempt. A low score means you have room to try Quality first and keep the picture closer to the source.
Why the resolution cap deserves more attention than it gets
The free plan caps the long edge at 768 pixels and paid plans at 1024, and if you need the higher ceiling the plans are on the pricing page. But the cap is not only a billing line. Several widely used detectors are unstable at high resolution: their scores track sharpness and JPEG artifacts more than they track how the image was actually made, which is also why they routinely call real photographs generated. Exporting at 1024px instead of the source size can improve the outcome for that reason, and it tends to make the SynthID stage more reliable as well.
When the export is still flagged
Work through it in order rather than rerunning the same settings. Confirm you are testing the exported file and not the original, which is the single most common cause. Switch to Strong if you were on Quality. Try the smaller output size. Then test against a different detector, because the first one may simply be wrong; the tools disagree constantly, and the reasons are set out in why detector bypass exists. If an image fails everywhere at every setting, it is a hard source, and a different frame will go through faster than a fifth attempt at this one.
What the number means, and what it does not.
Across three popular AI image detectors, internal testing puts the clean rate at about 90%. That number is ours, measured in house, and it is an average over a test set rather than a statement about your file.
Forensic Cleaner reduces detector-facing signals. It does not make an image undetectable and it makes no commitment about how any named detector or platform will behave, particularly after they retrain. Outcomes move with the picture, with which detector you test against, and with when that detector last changed.
Three things worth knowing before you upload.
- Embedded metadata does not survive the trip Validation converts your upload to lossless webp before any cleaning runs, and EXIF is lost in that conversion. That is a side effect of the pipeline rather than a dedicated metadata tool, but the practical result is that camera, timestamp, and location tags are gone from the export.
- Finished runs stay in your account until you clear them Each completed job is kept in your history, so you can fetch the file again, mark it, run it back through at another setting, or delete it. What is stored and for how long is written up in the privacy policy.
- Ownership is the line This is a tool for images you made or licensed. It is not for impersonation, for passing off someone else's work, for non-consensual imagery, or for anything otherwise illegal, and the safety rules say so plainly.
FAQ
Questions people ask mid-task.
What do I need before I can clean an image?
A PNG, JPG, or WEBP file that you made or hold a license to process, and an account. The free tier includes 10 credits a month and exports at up to 768px on the long edge. Anything in another format has to be converted before upload.
Should I run Strong or Quality?
Start on Strong. It reduces detector-facing signals harder, so it is the setting most likely to work on the first attempt. Move to Quality, which is on paid plans, only when Strong passed but changed the picture more than you can live with.
Why is my export smaller than the file I uploaded?
Output is capped at 768px on the long edge for free accounts and 1024px on paid plans, and the pipeline as a whole tops out at roughly 2 megapixels because of GPU cost. The cap is not purely a limitation: exporting at 1024px sometimes scores better against detectors that react to sharpness and compression.
The AI slop score came back high. Is it still worth running?
Usually yes. The score is an estimate of how hard this specific image will be to clean, not a prediction that the run will fail. A high score means you should expect to start on Strong, expect the cleaning effect to be more visible, and expect that more than one attempt may be needed.
The cleaned file is still being flagged. What do I try next?
Check first that you are testing the export and not the original, which is the most common cause. Then switch to Strong if you were on Quality, try the smaller output size, and test against a second detector. Detectors disagree with each other constantly, and several of them flag ordinary photographs.
Does this remove a SynthID watermark too?
SynthID removal is a separate optional stage rather than part of the main clean, so you have to turn it on. Medium strength is a sensible starting point, and the smaller output size tends to make that stage more reliable.
Can I run this over a batch, or from my own code?
There is no bulk upload in the web app. Pro and Scale include API access, so upload, validation, cleaning, status polling, download, and deletion can all run from your own scripts. How many jobs run at once depends on the plan.
Where to go from here.
Go deeper
- How to remove AI detectionThe same workflow, one control at a time.
- Bypass AI image detectionWhat detectors measure and where they break.
- SynthID removerThe optional watermark stage on its own terms.
- How to get rid of AIWhat signal reduction can and cannot do.
- AI detection removerThe tool overview, if you landed here first.
Free accounts get 10 credits a month at 768px. Starter, Pro, and Scale raise the cap to 1024px and unlock Quality mode, and Pro and Scale add API access.