Pricing
One balance across every generator and the API. Nothing is charged per user, per project or per seat — you buy credits a month and spend them on whatever you make, in the app or from your own code.
$1/ month
Free, plus API keys and MCP for your own code
$20/ month
Every model, and API keys to build with.
Everything in Starter, plus:
$20/ month
Every model, and API keys to build with.
Everything in Coding, plus:
$50/ month
Every model, and API keys to build with.
Unlimited on
For as long as you stay on Pro, while the offer runs
You pay $50 a month as usual — the place is taken by your first payment
Everything in Creator, plus:
$1/ month
Monthly only
Free, plus API keys and MCP for your own code
$19/ month, billed yearly
$228 once a year
Every model, and API keys to build with.
Everything in Starter, plus:
$19/ month, billed yearly
$228 once a year
Every model, and API keys to build with.
Everything in Coding, plus:
$45/ month, billed yearly
$540 once a year
Every model, and API keys to build with.
Unlimited on
For as long as you stay on Pro, while the offer runs
You pay $540 a year as usual — the place is taken by your first payment
Everything in Creator, plus:
Estimates
Compare plans
Our 3D generator and the catalogue's Hunyuan 3D models do not run for accounts in the 27 EU member states, the United Kingdom or South Korea. That is the Hunyuan licence, so it applies on every plan, Ultimate included.
A paid plan's credits arrive when each billing period's payment goes through, not on the first of the month. On a monthly plan, an upgrade is charged at once and adds the difference for the part of the period left when its payment goes through; a downgrade starts at the end of the period. On a yearly plan they arrive every month of the year, counted from the day you paid; an upgrade adds the difference for the rest of the current month, then the new plan's credits each month after, and a downgrade starts at the end of the year.
Working in someone else's project needs no paid plan: its people, items and canvases count against its owner's plan, and what you generate there is paid from your own credits. A lower plan deletes nothing.
The “at once” figures are for the app and your keys together. Through keys, at busy times the next job waits in the queue, or a catalogue model refuses it with lane_busy and charges nothing — through /openai/v1 a language model answers wait_capacity instead, with a Retry-After its client waits out.
Every plan
The plan decides how much you can spend, how fast, and which products open — each card above says which. It does not decide which guarantees hold or how a refusal is shaped: those are the same on every plan.
Generators for photos, music, video, 3D models, chat and code, each one an address you can POST to and all of them spending one balance of credits. There is also an app, on a computer or a phone, running the same generators on the same account, if you would rather click than call.
No. Every generator has an address you can POST to, and the app is one client of that same API rather than a separate product — both go through the same wallet and the same job records. Plenty of people never open the app after minting a key.
Every model on the Models page runs today, and the count there is live. GET /v1/models lists every address with whether it is available, and each has a page with its request shape and price.
No, and that is deliberate rather than unfinished. There are no CORS headers anywhere and a request carrying an Origin is refused. A secret key in a browser is a leaked key, and there is no browser-safe credential yet to offer instead. Call it from a server.
Nothing extra. Send the same idempotency key and you are charged once; the replay answers with the original job. A different key on the same account gets a 409 rather than your image — an idempotency key belongs to the key that first used it.
Only if you let us. Sign-up asks in a box of its own, separate from the Terms, whether you use an email address or a sign-in provider; it starts unticked, and leaving it empty costs nothing - the account is created all the same. An account that reaches the desktop without an answer is asked the same question once before it starts, with the same box, unticked, and nothing is used until it answers. The box used to start ticked, so an answer given that way no longer counts and the desktop asks again. Only what is made after you agree can be used, chat conversations only if you switch them on, and you can turn it off at any time in Account → Privacy, one category at a time. Once a category is off, nothing more of it is drawn into a training set — not even what you made while it was on — but what a training set already holds stays in it. The whole policy
Auray is free to open. Credits are only spent when you generate something, and they come out of one balance — in the app and through the API, on every plan.
