Two exciting new features are arriving at Google AI Studio, responding to a key request from developers. This platform for prototyping and experimenting with machine learning models now includes native video frame extraction and context caching. The video frame extraction feature is available today, while context caching will soon be added to the Gemini API.
With video frame extraction, developers can utilize videos uploaded to their applications, enabling Gemini to capture individual frames or image sequences. This functionality enhances the AI's ability to interpret scenes, generate concise summaries, and elevate user experiences. You can find this adjustable video frame extraction capability within the Gemini API.
Context caching allows developers managing large information sets to store frequently accessed context, which reduces costs and optimizes workflows. Rather than sending files to Gemini repeatedly, developers can send them once. Google highlights the utility of context caching for various scenarios, such as brainstorming content ideas, analyzing complex documents, and summarizing research papers and training materials. This feature will be supported in the Gemini API upon its release.
These updates are part of a series of exciting Gemini announcements at Google's developer conference, alongside the launch of Gemini 1.5 Flash, a new Gemma 2 model, and a pre-trained variant called PaliGemma.