This page does not rank products — naming or recommending commercial AI-NCII services is precisely what this site refuses to do. What it does is compare the categories the market consists of, on the dimensions that actually differentiate them: what they produce, what they require, and what exposure they create for the person using them.
| Category | What it produces | Input required | Convincing? | User's exposure |
|---|---|---|---|---|
| Inpainting "undress" tools | Real photo with clothing region synthesized away | One clothed photo | High — inherits the real photo's credibility | Upload is retained; UK: crime at creation; payment trace |
| Generation services | Wholly synthetic scenes with a target's face | Photo set or single reference (adapter) | Medium-high, with diffusion artifacts | Same legal exposure; often crypto-billed commissions |
| Face-swap video | Re-faced existing footage | Source video + one target photo | Highest for motion — original footage supplies realism | Hardest to run locally; web versions retain everything |
| Local workflows | Any of the above, self-hosted | GPU, open weights, forum knowledge | Varies with skill — enthusiasts get the best output | No operator risk; maximum legal risk (worst content made here) |
The same backend pipelines reach users through three delivery formats, and the delivery format — not the model — determines most of what the user actually experiences:
Upload-and-pay sites on disposable domains. The most accessible and the most dangerous to the user: card payments leave a traceable record tied to the crime, uploads are retained by anonymous operators, and the domain typically dies within weeks — often before a paid queue clears. A substantial fraction are pure scams that take payment and return nothing or generic output.
The same inpainting/generation backend behind a chat interface. Crypto or Telegram Stars billing keeps them alive longer than card-billed sites, and the channel outlives every domain. For the user the exposure is identical (uploaded photo retained, payment trail exists) with the added property that the operator is structurally unreachable.
ComfyUI-class node graphs on the user's own GPU. No intermediary sees anything — which is why the worst material, including CSAM, is produced here rather than through hosted services. Requires real technical effort; that effort is the only thing between the mainstream and the enthusiast tier.
Human operators running the pipelines above as a service: send the victim's photos, describe the scene, pay in crypto, receive custom output. The most directly victim-targeted format — production on demand against a named person — and treated accordingly by investigators.
People searching for alternatives to a given tool are really asking one of two questions, and the honest answer differs:
There is no answer this site will give — the commercial tier is a rotating cast of services that are legally indistinguishable (same exposure, same churn) and that this reference deliberately does not name. The technology differences above are real but secondary: every category processes a real person's likeness without consent, which is the thing the law and the harm actually attach to.
That has a real answer. The legitimate uses of the same generative technology — the ones that exist on stable domains with card billing because they don't get ejected — are: