I am getting an error when using Google Cloud Vertex AI to analyze images. This seems to happen for gif images, specifically. The error is: "the service reports an error with code INVALID_ARGUMENT described as: Request contains an invalid argument."

Is there any special situation related to animated gifs and the fileUri method of specifying a Cloud Storage bucket uri? The same code works well for non-gifs, and the same gifs work well when using Gemini outside of Vertex.

Full context:

• I am building a web app which allows users to upload images and have them automatically analyzed via Gemini.

• When using the "public" (external facing) Gemini API and an API key, I include the image as part of the request payload use base64-encoding. This has worked flawlessly for many months on several different image types, including gifs.

• When running in Google Cloud, I am trying to use Vertex AI to accomplish the same inference using the same prompt, but instead of inlining the image data I am passing a fileURI to my storage bucket. This works most to the time, but fails cryptically for some gifs.

• The same images which work for via the inlineData method (against the public API) do not work when using fileUri against the vertex AI.

• Model (same for both endpoints): gemini-3-flash-preview

An example gif that is failing for me is this one from the movie Point Break: https://giphy.com/gifs/break-point-ulvR63po0dMJ2

My gut is that because these gifs are "animated", there is some special scenario related to fileUri that I am missing. Perhaps they should be passed as video references somehow? Is there anything special with gifs and fileUri?

I can't find any reference to gif specifically in these Gemini docs. I can try to work on a repro, but it's a little tricky given that it references internal GCloud endpoints, buckets, etc.