I’m currently working with the Gemini AI API 1.5 Pro (latest version) and need to send large video files for inference. These videos are several hundred megabytes each (~700MB) but are within the API’s constraints (e.g., less than 1 hour in length). I want to upload them once and perform inference without re-uploading.

In GPT-4o, there was an option to use image_urls to reference images. Is there a similar method or best practice for handling large video files with the Gemini AI API 1.5 Pro?

The videos are too large to send repeatedly, so an efficient method for uploading and referencing them is crucial.

Any guidance on API endpoints, required parameters, or example code snippets would be greatly appreciated.

Thank you!