Video uploads to Gemini are getting so close now

Google’s Gemini continues to evolve, constantly becoming better at truly helpful tasks like summarizing content, answering complex questions, and even performing web research. However, a notable missing feature has been the inability to directly analyze uploaded video content. We’ve known Google has been working on bridging this gap since reports surfaced back in early March, and now we have the first real look at this video analysis capability in action.

The folks over at Android Authority managed to activate the feature within a beta version of the Gemini Android app (specifically version 16.15.38.sa.arm64), providing an early glimpse into how this functionality will operate. The demonstration reveals a straightforward process: users will simply hit the “+” upload button within the Gemini app interface and use the Android media picker to select video files stored locally on the device.

image credit: Android Authority

Once a video is uploaded, the real magic happens. Users can then pose questions directly related to the video’s content. Gemini processes the footage and provides relevant answers based on its analysis. The demo showcased impressive capabilities, including Gemini successfully identifying the geographic location depicted in a video without any additional context provided by the user. The system also capably handled the analysis of three separate video uploads within the same session.

There is one current limitation, though: the combined duration of all uploaded videos in a single session, or the length of a single uploaded video, is currently restricted to 5 minutes. Attempting to upload content exceeding this limit resulted in an error.

Despite this feature now appearing functional in beta, Google hasn’t officially announced a release date. With Google I/O just around the corner next month, it feels like a prime candidate for a full reveal, potentially highlighting it as a significant new addition to Gemini’s skill set.

Though I haven’t been able to test this out just yet, the potential applications are immediately apparent. Imagine quickly getting answers about a specific step in a lengthy tutorial video or having Gemini analyze a screen recording captured on your phone. It’s poised to be a genuinely useful enhancement to the already-capable AI, and I’m hopeful it arrives in an official capacity sooner rather than later.