Hey everyone!
I’m testing Gemini Pro Vision in Vertex AI and trying to have Google timestamp a video based on some events.
It seems to get the lenght of the video wrong, even if I only upload a video and ask for its lenght. Obviously it gets the timestamps wrong too.
Is this possible with the current model or is it too much to ask?
Thanks!
Hello edorainoldi,
Im also trying to get started with gemini vision pro with video as part of the prompt, but im unable to get the syntax on how to do it. I saw somewhere that the video has to be uploaded in the bucket inside the same gc project. Can you please confirm this and if possible can you share the basic syntax of how you send the video as the prompt?
Thanks
I am facing the same issue, literally asking it to identify the timestamp of an action in video and it’s getting it totally wrong, sometimes hallucinating timestamps which are clearly outside the video length!