I have extracted a set of 40 key images from a video. Can the 4o mini model accept 40 images, identify the top 5 images in accordance with the instructions, and provide me with the relevant images?
Is there another way to get the best images from the ones that have been provided?
Can it understand the provided images in sequence?
Hi!
For a number of reasons this would not be possible. You would fast exceed the token limit for gpt-4o-mini with that amount of pictures. Additionally, the model would struggle to analyze 40 pictures in a single API call.
If I was in your place, I would not provide more than 2 pictures for a given request; ask the model to return a description of the picture (or whatever it is you need as a basis to make a selection). Then combine all these outputs and run a final API call to make the selection of the top 5 pictures based on the relevant criteria.