Build with SpicyAPI

One uncensored API. Generative AI, straight into your product.

Start with production-ready image and video models, or a selected LLM. Compare content policies and price, test real parameters, then reuse the working request in your backend.

Browse models Pricing

How the pieces fit

From picking a model to carrying production traffic

01Pick the output you need — image, edit, video, or text — and read that page for the fields, limits and price dials of that task.
02Assemble the request on the model page, quote it, and confirm the estimate, the content boundary and the availability of that endpoint.
03Submit from your backend with an idempotency key, finish on the signed callback, and copy the result into your own storage.
04Set per-key spend caps, map the nine error codes onto product states, and only then raise the share of real traffic.

Everything below shares this

One envelope, one state machine, one currency

Every /api/v1 response arrives in the same envelope, {code, msg, data, request_id}, and every media task moves through the same five states: queued, running, succeeded, failed, expired. A failure carries one of nine stable errorCode values, content_rejected included; treat anything else as upstream_failed. Prices are quoted per published billing unit, with no credits in between.

One envelope, one state machine, one currency
Contract	Value
response envelope	{code, msg, data, request_id}
media task state	queued · running · succeeded · failed · expired
errorCode	9 stable values · closed set
pricing	Per published unit · no credits
availability	99.9% target, our own layer · not an SLA · excludes model providers · status.spicyapi.ai

Swipe horizontally to see all columns.

What you can build

One page per kind of output. Each one names the fields that move the price and the ones that make a result repeatable.

Video API solution

Turn a prompt or a still frame into finished video.

Compare video endpoints by duration, resolution, audio and price, then submit the same request from your server and track it by task ID.

Read guide →

Image API solution

Generate images with the controls the model really exposes.

Compare image endpoints by accepted inputs, size and seed controls, output count and per-image price, then reuse the exact request you tested.

Read guide →

Editing API solution

Edit the picture you already have, not one the model invented.

Send a picture you already have plus an instruction. How references get in, what the pixel limits are, and what editing endpoints charge for.

Read guide →

Integration guides

The mechanics underneath: getting a reference in, finishing a job after the request is gone, and speaking a protocol your SDK already knows.

Integration guide

Animate a still you already trust.

Turn a reference image into a clip. What first-frame quality, aspect mismatch, motion fields and duration do to the result and to the price.

Read guide →

Implementation guide

Finish the job after the request is gone.

Submit long media jobs from your server and finish them reliably: idempotency keys, signed callbacks, bounded polling and a five-state machine.

Read guide →

Integration guide

Keep your client. Change the base URL and the model value.

Point an existing OpenAI, Anthropic or Google client at SpicyAPI. What maps straight through, what is rejected on purpose, and how text is billed.

Read guide →

Moving from somewhere else

A contract-by-contract comparison for teams already calling another provider, built around a rollback you have actually tested.

Migrate one workflow, not a whole catalogue.

Move one workflow at a time: map fields, normalize states and errors, compare the bill, then shift traffic with a rollback path you have actually tested.

Read guide