Free · no upload · no account
AI metadata cleaner & remover
A combined AI metadata cleaner and remover: it reads every provenance and privacy field inside your image, shows you the real values, then removes what you choose, without re-encoding the picture.
Cleaning and removing are the same operation here. What differs from other tools is that you see what was found before anything is deleted, and the result is verified afterwards.
- Clean & verify in one pass
- Lossless, pixels unchanged
- Batch up to 30
Drop an image here
We read what is inside it first, then show you every field before anything is removed.
JPEG · PNG · WebP: up to 25.0 MB · 30 files at a time
Terminology
Cleaner, remover, stripper, what people actually mean
These terms get used interchangeably, and mostly they describe the same job: taking the hidden data out of an image file. Where they differ is emphasis, and it is worth knowing which one you need.
- AI metadata cleaner
- Usually means removing the markers that cause an AI label: C2PA manifests, XMP generation tags and PNG text chunks. The concern is how a platform treats your upload.
- AI metadata remover
- The same operation, named for the action rather than the outcome. Our AI tag remover page covers the generator-specific side of this.
- Metadata remover / stripper
- Broader: EXIF, GPS, device serial numbers and timestamps as well. If privacy rather than labelling is the goal, the full metadata remover page is the one you want.
- AI label remover
- Describes the outcome people are after, removing the thing that makes Instagram or TikTok attach a notice. Mechanically it is the same strip, which is why this tool covers all four.
How it works
Clean, then verify
- 1
Read before removing
Every field is parsed and listed with its real value. If you only want to look, the read-only checker does the same inspection without producing a file.
- 2
- 3
Strip without re-encoding
Metadata blocks are skipped on copy. The compressed pixel data is never decoded, so no quality is lost at any setting.
- 4
Verify the result
The output is parsed again and checked: pixel payload identical, file no larger, zero removable fields left. You see the outcome of that check, not a claim.
Where to go next
Pick the page that matches your problem
| If you want to… | Go to |
|---|---|
| Stop Instagram showing AI info | Instagram AI info guide |
| Fix a TikTok AI-generated label | TikTok guide |
| Remove GPS and device data | Metadata remover |
| Strip a Midjourney prompt and seed | Midjourney metadata |
| Clear Stable Diffusion PNG parameters | Stable Diffusion metadata |
| Just look, without changing anything | AI image checker |
| Understand a term | Metadata glossary |
Questions
Frequently asked questions
›What is an AI metadata cleaner?
A tool that removes the metadata AI generators and AI-assisted editors write into image files, C2PA content credentials, XMP generation tags, PNG text chunks and AI-related EXIF. It is the same operation as an AI metadata remover; the two names describe one job.
›Is this AI metadata cleaner & remover free?
Yes, with no account and no quota. The work runs on your own machine rather than a server, so there is nothing for us to meter.
›What is the difference between an AI label remover and a metadata remover?
An AI label remover targets the provenance markers that make platforms attach a notice. A general metadata remover also clears EXIF, GPS and device identifiers for privacy. This tool does both, and the mode switch decides how deep it goes.
›Does cleaning metadata guarantee the AI label disappears?
It removes the file-level causes, which is what most labels read. It cannot help when a platform uses a visual classifier or a pixel watermark such as SynthID, because those are not metadata at all.
›Can I clean images in bulk?
Up to 30 files per batch, repeated as often as you like. Each file is listed separately with its own findings, so you can see which ones carried markers.
References
Primary sources
The specifications this tool was built against. Where we describe how a format behaves, this is where that behaviour is defined.
- C2PA technical specification ↗
The provenance manifest format, including JUMBF embedding in JPEG and PNG.
- Adobe XMP specification ↗
The XML metadata block where generation parameters are usually written.
- IPTC Digital Source Type vocabulary ↗
Defines trainedAlgorithmicMedia and the other source-type values.
- W3C PNG specification ↗
Chunk layout, critical versus ancillary chunks, and the tEXt/zTXt/iTXt types.
- Google DeepMind SynthID ↗
The pixel-level watermark no metadata tool can remove.
Keep going