Warden Protocol

For images and videos, tokens work differently from normal text. The prompt you write is still converted into text tokens, but the generated image or video is not simply counted as (1 token = 1 word.[It could be less or more.] ) The cost usually depends on the model, resolution, duration and other processing requirements. For an image, the model processes your prompt as tokens and then generates the image based on that input. Some models also calculate the image itself using image specific units or compute, rather than ordinary text tokens. For video, the calculation can depend on factors such as resolution, duration, frame rate and the model being used. So you cannot directly calculate image or video cost by counting the words in the prompt.