Google Reportedly Working On a Conversation Summaries Feature for Gemini Live

Google is reportedly working on a new feature for Gemini Live, which will make it easier to track the conversations users have with the chatbot. As per the report, Google is developing a new conversation summaries button that will convert the two-way voice conversation into a comprehensive paragraph-style report with bullet points. The feature is said to appear once the conversation has ended, however, the implementation of the feature could be changed before the tool is released to the public.

Gemini Live's Conversation Summaries Feature

According to an Android Authority report, the Mountain View-based tech giant is working on the conversation summaries feature for the Gemini Live interface. The evidence of the feature was spotted by the publication during an Android application package (APK) teardown process within the Google app beta version 16.6.23. However, the feature is currently not live, and beta testers will not be able to try it out just yet.

Based on the screenshots shared by the publication, the new Gemini Live feature appears as a small box on the bottom-left corner of the Gemini app once a conversation has ended. In its current state, the feature just shows a few words of a prompt that can be shared with Gemini by tapping the box.

The box button sends a detailed prompt asking Gemini to act as a “highly skilled conversational AI assistant specialising in generating summaries,” as per the screenshots. The prompt also includes instructions for the summary such as “generate a brief, paragraph-style summary”, “create a non-generic title”, and not to include fluff. Once the prompt is shared with Gemini, it generates a comprehensive summary of the entire conversation.

As shown in the screenshots, it appears the feature is in its bare-bones form, and will likely not be shipped as such. Android Authority speculated that Google could either automate the process and automatically generate summaries after each conversation, or provide a button in the Gemini Live interface to generate the same. In either case, it is unlikely that users will still see the initial prompt, and it would likely become a backend process.