Uncensored creative tools
Add image and video generation to uncensored creative products. Choose models for your subject matter, quality, speed and budget.
Enterprise
Put production traffic on the API.Know exactly what you are signing.
Build creative tools, AI characters and private generation apps with uncensored image, video and language models. Use ready-to-integrate models, bring your own LoRA, or ask us to add a model the catalogue lacks.
At a glance
- One contracting entity
- The company that operates SpicyAPI. No reseller sits between you and the service.
- Priced in US dollars
- Per task, per model, with no credit or points system to convert.
- Spend caps on every key
- Daily and lifetime caps in the console, a monthly cap enforced on the API, plus separate workspace budgets.
Uncensored models are our focus
More room for the products you want to build.
SpicyAPI offers uncensored models without adding a platform-level content review. Choose models for images, video and conversations, then use your own interface and user accounts. We provide the API; customer content does not enter a public gallery.
AI characters and interaction
Choose uncensored language models for character chat, personalized interaction and storytelling, with images and video alongside them.
Private generation and custom brands
Build private tools for teams or customers. Tailor styles and characters with LoRAs you train and bring.
What we actually do
Three kinds of work, scoped and priced separately.
An enterprise engagement here is not a support tier with a badge on it. It is some combination of the three below, and the boundary between them is where the commercial terms get written.
- 01
Integration support
Getting your traffic onto the API
- Server-side key design: model allowlists, source-address limits and the spend caps that bound a compromised key.
- Asynchronous task handling end to end — submission, polling, signed callbacks and delivery of results from our own storage.
- Idempotency keys on write requests and a traceable identifier on every accepted task, so a retry never becomes a second charge.
- 02
Models on request
The model you need, on the same API
- Name the model or its Hugging Face repository. We check whether it can be served reliably, what it filters or refuses and what a call would cost before we commit.
- Once live it is an ordinary catalogue endpoint, on the same key, task API and billing as everything else.
- LoRA-ready bases take weights you train anywhere, passed by link on each request. We don't train or host them.
- 03
Commercial and settlement
How the money is handled
- Account and model-specific rates reviewed against your forecast volume and model mix.
- Postpaid credit by individual approval, with an agreed limit and expiry date that never becomes account balance.
- Balance movements, per-task charges and usage available in the console and exportable at any time.
Contract and settlement
How a charge is calculated, in plain terms.
Written commercial terms are agreed per engagement. Nothing on this page overrides the Terms of Service.
- Currency
- Prices, balances and records are all in US dollars. There is no credit or points layer to convert between.
- Before the call
- The console shows an estimated cost for the exact parameters of a request before it is submitted.
- How a task settles
- An accepted task reserves funds first and settles against the real result. If the settled cost is higher than the reservation, the difference is absorbed on our side rather than billed to you.
- Spend controls
- Every key carries a daily and a lifetime cap you set in the console, and a monthly cap enforced on the API. Workspaces carry their own budgets, and a reservation that has not settled still occupies the limit.
- Postpaid credit
- Credit is off by default and granted only by individual approval. Approved headroom is not balance, it expires on the agreed date, and unpaid or exceeded credit blocks new tasks.
- Invoices and statements
- Each completed top-up can be issued as a payment invoice, and each month as a usage statement of what was actually spent; both download as PDF from the console. The billing name, address and an optional tax number are yours to enter and change there, and a document already issued keeps the details that applied when it was created.
Top up by card, crypto or transfer. Usage priced in USD.
Card payments are credited immediately, and a crypto payment (USDT included) is credited once the network confirms it. A manual transfer is credited once we verify the reference you submit, usually within a few hours; the console shows which methods are enabled on your account and the exact payment details. For larger volumes, discuss account or model-specific rates, or apply for individually approved postpaid credit.
Your content stays out of the public gallery.
Inputs, prompts and results are used to fulfill your requests. We do not publish customer content on our website, in a gallery or as a showcase.
Read the privacy policy
No public customer content
There is no public creation feed. We do not publish your business assets or generation results.
No training without permission
We do not use your inputs or outputs for model training.
Control keys and access
Authenticate with API keys, limit a key to named models and source IPs, and cap what it can spend by day, by month and for its lifetime. Keys are stored as hashes and can never be shown again, and any key can be revoked immediately.
Defined retention periods
Uploads are kept for 1 day, generated media for 14 days, prompt text is erased after 30 days, and network metadata is pseudonymised after 90 days. Billing records and legal preservation exceptions follow the privacy policy.
- 1day
- Uploads
- 14days
- Generated media
- 30days
- Prompt text
- 90days
- Network metadata
A clear path from requirements to launch
- 01
Share your goals
Describe your use case, models, expected volume and existing integration.
Your side
Send the models, workloads, forecast volume and the integration you already have.
Our side
Read the inquiry and confirm which parts the published catalogue already covers and which need work.
- 02
Evaluate the fit
Agree on model quality, data requirements, costs and acceptance criteria.
Your side
Provide acceptance examples and the constraints that decide whether a result is usable.
Our side
Assess model fit, data requirements, cost per task and any commitment we can put in writing.
- 03
Validate, then scale
Verify the agreed scope before progressing to integration and broader rollout.
Your side
Run the agreed scope against real traffic, with the spend caps still in place while you do.
Our side
Work through the failures that only appear at volume, then agree the rollout and the terms.
Operating targets
Targets we publish, and the SLA we do not sell.
Neither number is a contractual SLA, and a missed target earns no service credits, refund or other compensation. Live component status is published at status.spicyapi.ai rather than summarised into these figures.
- 99.9%
- Operational target for our own layerRequest intake, task intake and callback delivery.
- 1business day
- Target for the first reply to a support requestCounted from the moment your message reaches us. It goes to the people who operate the service, not to a queue that only forwards it.
Outside the availability target
- The third-party model services that execute a generation.
- Maintenance windows announced in advance.
- Events outside our control.
Where a service level is signed separately, that document governs for the parties who signed it. This page publishes targets; it does not replace an agreement.
Procurement questions
The questions that decide a review.
- Who are we contracting with?
- The company that operates SpicyAPI, trading as spicyapi. There is no reseller in between: the party you contract with runs the platform and answers support.
- Can you issue an invoice, and how is VAT handled?
- Yes. Every completed top-up can be issued as a payment invoice and every month as a usage statement, both downloadable as PDF from the console and both carrying the billing name, address and optional tax number you enter yourself. Tax is shown as zero: we are a US supplier, we collect and remit no VAT, GST or sales tax on your behalf, and each invoice states that the customer self-assesses under the reverse-charge rules of its own jurisdiction. We give no tax advice, and no document we issue is certified or registered with any tax authority.
- Is our data used to train models?
- No. Prompts, uploads and outputs from everyday API calls are not used for training, and results never enter a public feed.
- How long are generated files kept?
- Uploads are kept for 1 day, generated media for 14 days, prompt text is erased after 30 days, and network metadata is pseudonymised after 90 days. The trust centre lists every retention window in one table.
- Do you hold a compliance certification or offer an uptime SLA?
- No certification, and no contractual SLA. What we publish is an operating target: 99.9% for our own layer — request intake, task intake and callback delivery. It excludes the third-party model services that execute a generation, maintenance windows announced in advance and events outside our control, and a missed target earns no service credits, refund or other compensation. Live component status is at status.spicyapi.ai. Where a service level is signed separately, that document governs for the parties who signed it. We claim no SOC 2, ISO 27001, PCI DSS or HIPAA eligibility.
- How do we cap what a compromised key can spend?
- Each key carries a daily and a lifetime spend cap you set in the console, plus a monthly cap enforced on the API. A key can also be limited to selected models and source addresses, and revoked immediately. Workspaces carry separate budgets on top.
Enterprise
Bring uncensored models into your product
Tell us your content types, model needs and expected usage, including any model you need that the catalogue lacks. We will work through model selection, integration and pricing with you.
The form below reaches authorised staff of the company that operates SpicyAPI, the same company you would contract with.
Who operates this service
SpicyAPI is run directly, not resold.
One company stands behind the service, the terms and the money. There is no reseller in between: the party you contract with is the one that runs the platform and answers support.
The registered address and company number are supplied on request during contracting. Nothing on this page is a certification, and we do not present a supplier's certification as our own.
- Business address
- 2261 Market Street STE 72933, San Francisco, CA 94114, United States
- Trading name
- spicyapi, at spicyapi.ai
- Billing currency
- US dollars, charged per run. No credit or points system.
Documents and channels