Fair use

What the limits actually are

Most services put a page like this up to explain what “unlimited” really means. Ours says the opposite: nothing here is unlimited, and nothing is secretly throttled either. You buy credits, a generation costs credits, and the price is shown before you press the button.

There is no unlimited tier

Every model on Katama is credit-metered on every plan. No plan grants unmetered generations, and no plan quietly slows you down once you have used “too much”.

We think that is the honest arrangement. An unlimited plan with a hidden throttle is a worse deal than a metered one with a visible price, because you only find the ceiling by hitting it.

What is limited: how FAST, not how MUCH

The only real limits are request rates, and they exist to stop scripts, not people. You will not meet them by working normally — they are set well above what a person clicking through a studio can reach.

ActionBurstSustained
Generate (image, video, audio)12 in a row~720 / hour
Run a full pipeline6 in a row~360 / hour
Improve a prompt30 in a row~1800 / hour

“Burst” is how many you can fire back to back; “sustained” is roughly how many per hour once the burst is spent.

Uploads

Uploads are counted against a rolling 20 GB per 30 days, and at most 2,000 files in that window. Rolling, not calendar — so a month boundary is not a reset you can plan around.

This is a storage ceiling, not a work ceiling. It is set high enough that normal production never reaches it; if yours does, tell us and we will raise it rather than leave you stuck.

What is not allowed

Automated bulk usage — scripts, bots, or resale of raw generation capacity — is not covered by a personal plan. If you need volume through an API, that is what the developer API and reseller terms are for.

What happens if you hit a limit

You get a clear 429 and a retry hint. Nothing is deleted, no credits are taken, and your account is not suspended for hitting a rate limit. A failed generation is never charged — the hold is released.

The numbers on this page are read directly from the running configuration, so they cannot drift from what the service actually enforces. See also pricing and terms.