SpicyAPI

SpicyAPI is an application programming interface for generating adult images, video, speech and chat. Its home page calls it the porn API built on frontier models, and it names these uses: AI companion apps, creator platforms, roleplay chat and adult marketplaces. The terms name the operator as Wayfinity Ltd, a company incorporated in the Republic of Cyprus and trading as SpicyAPI. The terms say the service has been at its domain since June 2025. The buyer is a developer or product team, not a site visitor. The pages are written for an operator who wants to put generation inside its own app and bill its own users.

What it sells an operator

The product is one API key and one base URL for every model. The pages group the models into image, video, speech, live voice, transcription, embeddings and chat. The main image models are Spicy Image 1 Pro, Spicy Image Photo 1 and Spicy Image 1, priced by the image, with Spicy Image Action 1 for ready-made scenes, Spicy Image Reference 1 for generating from a picture, Spicy Image Edit 1 and Spicy Upscale 1. In October 2026 the pricing page lists Spicy Image 1 Pro at $0.09 an image.

The video range is the largest part of the catalog. On the pricing page read in October 2026, Spicy Motion 3 and its faster variant handle image and text to video with native sound, at 480P, 720P and 1080P output and clips of 2 to 30 seconds. The pricing page lists Motion 3 from $0.10 a second at 480P, with higher resolutions priced above that. Other video models include Spicy Cinema 1, a cinematic engine with output from 480P to 1080P and character references, Spicy Motion 2 for lip-sync, Spicy Character Video 1, a silent Spicy Motion Draft tier for checking motion before a final render, Spicy Video Edit 1 and Spicy Animate 1 for motion transfer. A 60 fps option adds 20 percent to the video price, and the vendor refunds that fee if the step fails.

Speech comes in several voice models with inline tags such as whispers and laughs, and custom voices can be made from a clip of a speaker who has consented. Live voice calls run over a WebSocket, with a session limit of 13 minutes per call according to the docs. Chat models include Spicy Companion 1, which the docs describe as the recommended model for explicit roleplay, and a lower-priced Companion Flash tier. The companion model is listed in the same October 2026 pricing at $1.00 per 1M input tokens and $2.80 per 1M output tokens. It can return up to 4 alternative replies per request.

Limits are stated in the docs. Image prompts are capped at 4000 characters, outputs run to 6 images per call, and every account starts at 600 requests a minute, which the FAQ says is raised automatically toward 6,000 when an account keeps reaching it. Up to 20 video jobs run at a time, raised on request.

The terms that decide the choice

The terms describe the service as generating adult content, including explicit sexual imagery, and say it is designed for platforms and applications "serving adult content to verified adult users." They say the services are intended for business and commercial use, and the terms open with an adult content warning that the service is only for users 18 years of age or older.

The acceptable use policy is strict on people. It says images and video must not depict, edit, animate or import a real, identifiable person, living or dead, and that there is no consent route for images or video. Voice cloning is allowed only from a recording of a speaker who consented. The policy says the company screens prompts and withholds output that shows a recognizable person. It also sets a zero tolerance rule for violations involving minors, content made or shared without the consent of the people shown, or illegal activity, with immediate account termination, a permanent ban and potential reporting to law enforcement. The listed penalties also include forfeiting the account balance. Other enforcement decisions can be appealed within 14 days by email, but zero tolerance cases cannot.

Inputs are restricted too. The terms say the services do not accept uploaded images or video, with two exceptions. A picture sent to the reference image model is recreated as different, fictional people, and a character import feature is enabled on request under a separate schedule at the company's sole discretion. The terms call that import feature not part of the standard services. Image-to-video and image-edit work only on images the same account generated. A site that plans to start from its own photography or footage will not find that route on the pages.

Age checks run at two levels. The acceptable use policy says that where the law of the account holder's location requires it, the account holder completes a one-time age verification with the company's provider, VerifyMy, after the first top-up and before the account can generate or retrieve content. The policy lists the places this applies: the United Kingdom, France, Germany, Italy, Australia, Brazil and a long list of US states, and it says the list may change as laws change. For end users, the operator must verify that everyone is 18 or over before they reach generated content, follow the rules of each country it serves, and keep audit logs. The policy says it does not prescribe a method, that a self-declared age is not enough where the law requires verification, and that the company may audit an operator's age gate and end the service if it falls short.

Money and refunds carry several rules. Credits are prepaid with a minimum top-up, paid by Visa, Mastercard or Apple Pay, or by BTC, ETH or USDT. Balances never expire, and failed or withheld generations are refunded. In October 2026 the pricing page says bonus credit comes with larger top-ups: a $500 top-up gets $100 extra and a $1,000 top-up gets $300 extra. The terms say unused balance from a card top-up can be refunded to the card that paid, on request. Bonus credit, crypto top-ups and character import fees are non-refundable, except where the law requires otherwise. The pages state no platform fee, no egress charge and no monthly minimum, and they say price changes come with at least 14 days of email notice.

Output rights are the clause most operators will read twice. The terms grant a worldwide, perpetual, non-exclusive, royalty-free license to use and distribute generated output for any legal purpose, including reselling it and selling credits or paid access built on the services to the operator's own users. They also say the company does not use inputs or outputs to train AI models. The terms add that outputs may not be unique, because other users may generate similar content.

Storage is set out as a fair use rule. Outputs stay available to the account at no extra charge. The output URLs are for retrieval and reuse as inputs, not for serving an operator's traffic, so the operator must download outputs and serve them from its own infrastructure. The docs say the same thing in other words: the CDN copy is a master copy and not a streaming host.

Two more points concern record-keeping and labeling. Images are screened before and after generation, and the file metadata of every image and video carries a machine-readable AI label, which the vendor says the EU AI Act requires. The 2257 statement says all visual content produced through the service is wholly computer-generated, that no real person may be used as an input, and that the record-keeping requirements of 18 U.S.C. 2257 do not apply to content generated through the service. The statement places responsibility on any customer that combines generated content with real performers' content in its own products. That is the vendor's stated position.

Complaints go to one address. The terms say copyright claims are handled under a separate DMCA Policy and that the company keeps a repeat infringer policy, and the acceptable use policy sends abuse reports, copyright claims and law enforcement requests to the same contact email; the 2257 statement points to a report form for content that appears to show a real person. The terms also make the operator solely responsible for moderation, age verification and safety measures when it serves generated content to its own users. The self-serve account has no fixed term: the operator can close it at any time, the company gives at least 14 days of email notice before retiring a model, and the terms are governed by the law of England and Wales.

Getting started and support

The route in is self-serve. The docs describe signing up, adding credits, creating a key and sending a first request, and the home page says most teams ship a first generation in about 10 minutes. Keys can be created per project or per agent, with sandbox keys for staging, and revoked in one click from the dashboard. The base URL is plain HTTPS, the chat and embeddings endpoints follow the OpenAI request shape, and the docs mention an MCP server for agents.

Video requests are asynchronous. They return a task that can be polled or delivered by webhook. Webhook calls carry an HMAC-SHA256 signature over the raw body, must be answered with a 2xx within 8 seconds, and are retried up to 6 times, after 1 minute, 5 minutes, 15 minutes, 1 hour and 4 hours. Image, speech, transcription, embedding and chat calls are synchronous. The request log shows cost and latency for each call, and a usage view allows reconciling spend day by day.

Larger accounts can move to an Enterprise contract, which the pricing page describes as negotiated rates, higher limits, dedicated support with an SLA, and custom billing or invoicing. Bank transfer is offered for larger accounts by email. The pages state no support hours and no response time for the self-serve tier. The support contact is an email address on the terms, and an operator who needs a written SLA must ask for an Enterprise contract.

Company and track record

The company is Wayfinity Ltd, and the terms name it as the operator of the service. The terms were last updated on 8 October 2026. They say the service is not connected with any other service using a similar name. The pages name no team members, no office address beyond the country of incorporation, and no client list. The home page cites a 99.9% uptime figure and a 12 second median generation time, which the vendor presents as its own claims. Neither figure comes with a measurement method on the pages.

Who it suits

SpicyAPI licenses generated adult content to operators under written output rights, and the vendor's own description covers AI companion apps, creator platforms, roleplay chat and adult marketplaces.

It fits a developer team building a companion or roleplay product that needs pay-per-unit pricing, a chat model that holds a persona, and a clear output license that allows the operator to resell generated content. It fits a marketplace that wants generated stills and short clips made from text prompts, with no uploads of real people, and it fits an operator that wants to pay by card or crypto with no monthly commitment.

It fits less well a site that needs to start from its own photography or footage, because the pages restrict uploads. It fits less well a business that needs a fixed subscription, because the pages say there are no plans. The operator carries the moderation and end-user age checks for its own product. Bonus credit and crypto top-ups are not refundable, and the age verification step adds an onboarding gate that will vary by location.

The strength is clear and specific. Unit prices are published in US dollars, failed jobs are not billed, output rights are written out in full, and the API is documented model by model, down to the sizes, durations and limits. For a site operator who needs a generation layer with stated terms, the published pricing, policy and documentation give a full basis for the decision before any money is spent.