Billing Details for Image Processing in Multi-Turn Conversations with GPT-4V API

I’m seeking clarification on the billing strategy for image processing in multi-turn conversations via the GPT-4 API. Specifically, I’d like to understand if an image is reprocessed and billed for each message in a conversation where the chat history includes an image.

Given that the chat completion API doesn’t maintain state across turns, it seems that re-uploading the image with each turn could lead to multiple charges for the same image. This could become cost-prohibitive for applications that need to reference an image across multiple interaction steps.

Could anyone with experience on this topic inform us whether each image reprocessing incurs an additional fee, or if there are billing considerations to mitigate this in multi-turn scenarios?