Solved: Re: Gemini video analysis fps control - Google Cloud Community
From what I can see, it's not possible to control the fps rate at which gemini models analyze video. My use case is the analysis of a very short (4-5 second) video with fast movement. This doesn't seem possible at the moment. Does anyone know if this is a feature that would be supported in future? I'm thinking that as a workaround I'm going to have to rather extract frames myself from the video and feed them in as a sequence of images. Is that the correct approach?
Frame Extraction: By extracting frames from the video at a higher FPS (e.g., 30, 60, or more frames per second), you give the model more data to work with, allowing it to better analyze fast movements.
Sequence of Images: Once you’ve extracted frames, you can input them as a sequence of images into the model for analysis. This way, you have control over how many frames you feed into the model, and you can optimize for the specific speed of movement in the video.
Speed & Efficiency: If you extract frames at a higher rate than the default FPS, the analysis might take longer, depending on the number of frames you extract, but it should give you better results for fast movements.
For more information, you can explore these Gemini models and their video analysis capabilities. Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.
Hi @DeanGracey, Welcome to Google Cloud Community! You're correct that currently, the Gemini models (or similar models) don’t offer direct control over the FPS rate when analyzing video. Typically, these models analyze the entire video as a whole, without granular control over the frame rate, which can be an issue for fast-moving videos where higher FPS is needed to capture more detail in quick actions. The workaround of extracting frames yourself is a solid approach.
View solution in original post
