The rate card

Billed per output image. A range means the rate depends on resolution or quality.

Billed on the tokens the prompt and the image weigh, so cost scales with the resolution asked for.

Billed per second of output video — the rate depends on resolution and aspect ratio, and a clip costs its duration times that rate.

Billed once per generated clip, whatever its length — the rate depends on resolution and frame interpolation.

Billed per token, so a run costs what its prompt and answer weigh.

Billed per second of input audio, whatever the transcript weighs.

Billed per thousand characters of the text spoken.

Billed on the tokens the text and the spoken audio weigh, rather than on the characters submitted.

Billed once per generated file, whatever its length.

Billed for the seconds the model runs — cost scales with resolution, settings, and input size.