Loading...
Loading...
18+ check appears when you click Generate
Most pages ranking for “AI undressing” open with “cutting-edge AI” and a wall of five-star testimonials nobody verified. This one doesn't. We run the tool ourselves. What follows is the queue's real numbers, the failure modes we see in support tickets, the 2026 legal picture in your country, and what 3 credits actually buys.
If you're here to harm someone, this isn't the site. Read the “What we won't generate” section before uploading anything — the gate is a real one, not a checkbox.
3 credits per run. Only the result is yours to keep — your upload is deleted after the task finishes.
Real output from this tool. Front-facing portraits with even lighting work best.
Drop your photo here
JPG / PNG / WebP, up to 10MB
A 3-question 18+ check appeared when you uploaded. Re-open it from the locked button if you dismissed it.
Side-by-sides from the queue, picked roughly at random. We didn't cherry-pick the best ones — these are the kind of result a typical front-facing portrait gives you.
A latent-diffusion inpainting pipeline running on a 48GB GPU. You upload one photo of yourself (or someone who has agreed to this in advance). The model masks the clothing region of the body it detects, samples plausible underlying anatomy from its training prior, and re-renders that region back into the original lighting.
That's the whole AI undressing pipeline. There's no “X-ray” happening. The model is guessing. Sometimes that guess looks convincing, sometimes it looks wrong — see the failure gallery below.
One run takes around 50–80 seconds in our queue depending on load, and costs 3 credits (roughly 30 cents). New accounts get free starter credits so you can see what the output looks like before topping up.
This is the question almost every page on this keyword refuses to answer. The honest version is: legality of AI undressing depends on whether the subject consented, whether the result is shared, and where you live. Here is the 2026 picture, condensed.
United States
The federal TAKE IT DOWN Act (signed May 2025) criminalizes publishing non-consensual intimate imagery, including AI-generated. Sharing without consent is a federal offense. Generating for personal viewing is currently not federally criminalized in most states, but states like Minnesota, Texas, Virginia, and California have their own statutes that do reach generation. Check your state.
United Kingdom
The Online Safety Act 2023 made sharing deepfake intimate images an offence. The 2024 amendment under consideration extends this to creation regardless of distribution. If you're in the UK, treat AI undressing itself as legally risky.
European Union
The EU AI Act classifies non-consensual deepfakes as prohibited. GDPR adds that the person depicted has rights over their image. Result: even with consent, you should keep written proof of it.
South Korea, Australia, Canada
All three have introduced or strengthened deepfake-specific laws since 2024. Korea's amended Sexual Violence Act now criminalizes possession with intent to view. Australia's Criminal Code Amendment (Deepfake) Bill 2024 covers creation. Canada is following provincial routes.
None of this is legal advice. We are not lawyers. If your use case is anywhere near the line, talk to one before uploading anything.
Five yes/no questions. If you can't answer all of them comfortably, AI undressing isn't the right fit for what you're trying to do.
Knowing when not to use the tool will save you credits. These categories of input either return obviously broken output or get rejected by our pre-checks. None of them are bugs we're trying to fix — they're fundamental to how diffusion inpainting works.
Group photos
The model picks one body and gets confused. With more than one person, segmentation drifts.
Heavy occlusion
Crossed arms, bags, oversized jackets — anything covering more than ~40% of the torso. The prior has nothing to anchor on.
Extreme angles
Side-back views, bird's-eye, fisheye lenses. Trained primarily on front-facing portraits.
Anime, 3D renders, paintings
Real-photo prior. Stylized inputs come back uncanny — neither the original style nor a clean photo.
Very dark or low-resolution photos
Below ~512px on the short side or ISO-noise-heavy phone shots: output looks washed.
Public figures we have on a blocklist
If our face-recognition pre-check matches a flagged identity, the request is rejected and credits are not charged.
We get asked for these features fairly often. The answer is no, and we're not going to be talked into it.
A lot of pages on this keyword bury pricing or hide it behind a sign-up wall. Ours:
| Item | Credits | ≈ USD |
|---|---|---|
| Free starter (new accounts) | 20 | $0.00 |
| One generation | 3 | ~$0.30 |
| 100-credit pack | 100 | $9.99 |
| 500-credit pack (recommended) | 500 | $29.99 |
| Failed generation | 0 | Auto-refunded |
Pay-as-you-go. No subscription, no auto-renew. Sign in via Google, top up with Stripe, that's the whole loop.
If you came here for a different reason —
We have a separate page positioned for designers and e-commerce teams. Same underlying model, but the workflow, pricing examples, and showcases are tailored for outfit replacement rather than explicit output.
Open AI Clothes RemoverAll running on the same VideoFaceSwap account and credits.