The clean image API and workspace
Remove any watermark.
Hidden signals. Provenance metadata. Tiled patterns. Branded logos. One engine handles all of it.
No credit card. No result, no charge. Cancel anytime.
Trusted by teams processing images at scale
The problem
Watermarks are everywhere now. And they are not going away.
Every major image platform stamps its work. Some signals are visible, branded into the corner of a stock photo. Others are hidden in the pixels or buried in the metadata. Either way, the image you downloaded is not really yours to use cleanly.
Hidden watermarks
Pixel-level signals like SynthID are baked into the image data. You cannot see them. You cannot crop them out. They survive resizing, screenshots, and re-encoding.
Model-embedded signals
Stable Signature and similar signals are baked into a generative model's decoder itself. Every image that model emits carries the mark, invisibly, by default.
Frame-level marks
Meta Video Seal and similar frame watermarks tag generated or altered video frame by frame to trace origin. The signal rides along every extracted still.
Provenance metadata
C2PA manifests and EXIF provenance fields travel with the file. They declare who made it, when, and with what tool. Useful for some workflows. A liability for others.
Tiled patterns
Repeating marks tiled diagonally across an image. Stock libraries and preview downloads use them to ruin the file until you pay. Even after paying, the tile pattern often stays.
Branded logos
The classic corner watermark. A platform logo stamped over the image. Easy to see, historically hard to remove cleanly without leaving artifacts or blurring the area behind it.
One engine handles all six. Without artifacts. Without metadata residue. Without per-image fees that punish scale.
How it works
Three steps. From watermarked to clean.
Same flow whether you process one image or ten thousand. Same engine whether you use the API or the Workspace.
Send your image
Upload one image or a folder of them in the Workspace. Or POST it to the API from your pipeline. Single endpoint, single or bulk.
We strip the signal
The engine detects every watermark class in the file. Hidden signals, provenance metadata, patterns, logos. Removes them without touching the image content you want to keep.
Get the clean file
Download the result. One file at a time, or the whole batch as a ZIP. Every job lands in your archive, searchable, for whenever you need it again.
For developers
A watermark removal API that behaves like the rest of your stack.
One endpoint. Single image or bulk, same call. Async by design — submit a job, poll its status and queue position, then fetch the clean result.
Generate a key in the Workspace. Cap spend per key. Set rate limits per key. Everything you would expect from infrastructure you depend on in production.
Token-based pricing means the math is predictable. 10 tokens per image. Always. No surprise invoices. No per-seat licensing.
{
"jobId": "job_abc123",
"status": "queued",
"tokensCharged": 10,
"queuePosition": 2,
"poll": "/api/jobs/job_abc123",
"result": "/api/jobs/job_abc123/result"
}For everyone else
The Workspace. Built for bulk, designed for humans.
Drop a folder of images. Watch them process in parallel. Download the clean set as a ZIP in one click.
Every job lands in your archive. Searchable by filename, date, and watermark class. Pull the clean originals back any time a client asks.
No per-seat pricing. No software to install. No learning curve that costs you a Monday morning.
Everything in one account
Two surfaces. One set of capabilities. Nothing gated behind a sales call.
What we remove
Six classes of watermark. One engine.
If it is a signal on or in your image, we strip it. Here is the full list, with no marketing fog around it.
AIUnmark is for images you own or have the rights to clean. Archive restoration, client deliverables, catalog normalization, dataset preparation.
Pricing
Token-based. Predictable. Cheaper as you scale.
One unit: 10 tokens per image, always. The only thing that changes between plans is how cheap each token gets.
Trying it on real work
Most popularRegular weekly volume
Studio and team pipelines
Catalog-scale processing
Every account starts with 5 free images. No card required.
From the people using it
Built for the work, not for the screenshot.
Developers, studios, and catalog teams running AIUnmark in production.
I dropped the API into our media pipeline on a Friday afternoon and forgot about it. Three weeks later it had quietly processed 40,000 images. The spend cap meant I never had to explain an invoice to anyone.
We used to lose a half day per client to manual cleanup. Now we upload the folder, click ZIP, and the deliverable is done before the standup. The archive means when a client comes back six months later, the results are still there.
Our cost per processed image dropped from $0.40 on our old tool to under a dollar at scale. That is the difference between processing a catalog and not bothering. We process the catalog now.
Questions
The things you want to know before you sign up.
Short answers. No marketing fog. If a question is missing, the contact form is at the bottom of the page.
Still have a question? Talk to us.
Ready when you are
Remove the watermark. Keep your image.
Five images on us. No card. No call. No risk. If it doesn't work, you don't pay.