Get started

Test mode

A test key (rt_test_…) runs the full lifecycle of a generation without charging anything. It is meant for development and continuous integration.

	Test key	Live key
Prefix	`rt_test_`	`rt_live_`
Result	Watermarked sample image, in about 3 seconds	Real generation
Charge	None, never a `402`	Price of each accepted generation
`livemode`	`false`	`true`
Instruction moderation	No	Yes
Heavy load refusal (`503 overloaded`)	No	Render, edit, decoration

Same checks as a live key: permissions, parameters, images and masks, rate limits and errors, 503 mode_unavailable included. Received images are checked, then deleted as in live mode.
Sample image: a watermarked gradient in WebP, with the proportions of the input image, 1,024 pixels on its longest side, whatever the resolution (quality) or factor (scale) requested.
Same statuses as in live mode: the generation stays in_progress for about 3 seconds and can be canceled during that time.
cost gives the price the generation would have in live mode.
Separate modes: a key only sees the generations of its mode; a generation of the other mode responds 404 resource_not_found. Idempotency keys and the quota of generations launched are specific to each mode; the requests per minute limit is shared by both.

Moving to live mode

Top up the balance from the console (see Billing).
Create a live key with the permissions you need.
Replace the test key with the live key on the production server: the code does not change.