Google adds app links to AI Mode and selfie avatars to Vids

Google integrated its AI Mode search into external apps on Thursday. It also allowed users of its Vids tool to create video avatars from a selfie. The two updates deepen the company’s conflict with OpenAI and Anthropic over who owns everyday AI tasks.

The first change connects AI Mode to a basic set of third-party services, including Instacart, Canva, and YouTube. AI Mode used to simply answer questions within Google’s conversational search. Linking an account now converts those responses into actions.

Google’s example shows a user creating a grocery list for a barbecue in AI Mode and then sending the ingredients directly to an Instacart cart for checkout. A separate demo shows AI Mode creating a party playlist and saving it to YouTube Music. Canva is used to create templates for flyers.

The rollout is currently limited to the United States. Google stated that it is expanding its partnerships and will soon support additional apps.

Google is primarily concerned about retention. App links encourage users to complete planning and shopping tasks through AI Mode, which is better than opening a search page and leaving. It also closes the gap with competitors.

OpenAI’s ChatGPT and Anthropic’s Claude already allow external apps to communicate with their assistants. The move builds on what Google demonstrated at its I/O developer conference earlier this year, when it allowed people to connect apps such as Canva, OpenTable, Spark, and Instacart to the Gemini app to complete tasks faster.

Google builds a digital stand-in for Vids

The second announcement focuses on video. Google is updating Vids so that users can create digital avatars that look and sound like themselves. It’s created using an uploaded selfie and a voice recording. Vids began as a tool for AI-assisted work presentations. The avatar feature, combined with new editing tools, positions it as a general video creation platform.

Google is also integrating Gemini Omni into Vids. Omni represents the company’s multi-modal model. It combines a written prompt and reference images into a finished clip. Omni can change the background, adjust the lighting in phone footage, and add effects. It now supports step-by-step edits, allowing users to adjust a video without restarting.

The company included guardrails on the avatars. Each one relates to the account holder’s likeness and Google account. Each carries an invisible SynthID watermark, too. In some regions, access remains restricted to users aged 18 and up.

Those limitations appear to be a reaction to how synthetic video has previously failed elsewhere. OpenAI’s Sora, which has since been discontinued, allowed users to create videos of public figures, including OpenAI CEO Sam Altman. Vids’ personalized avatars and conversational editing now directly compete with video AI companies like HeyGen, Synthesia, Captions, and D-ID.

Cryptopolitan reported this month that Google Images is being rebuilt into a Pinterest-style “For You” gallery. It’s also incorporating AI image generation into Search using its Nano Banana model.

According to Cryptopolitan, Google DeepMind CEO Demis Hassabis has also called for a US-led body to oversee AI. During the same period, publishers filed a class action lawsuit. The plaintiffs, including Hachette Book Group, Cengage Learning, and Elsevier, allege that Google trained its Gemini models on copyrighted books without permission.

If you're reading this, you’re already ahead. Stay there with our newsletter.