Question

Hi,

I noticed that gemini supports additional arguments such as trimming video and downscaling media for cheaper processing -https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/video-understanding#customize-video-processing

These two things would definitely be useful, but I haven't noticed any option to use them with PydanticAI, I could be wrong as I'm just learning this library, but I would really appreciate an answer.

Basically, having such an option would reduce the cost of using gemini for video processing by 3 times, which is quite a significant amount + would allow us to hard limit the token usage per video (for example by analysing only the first 5 minutes of the video), which is another useful thing.

Best regards

Additional Context

No response