What Are Google's Biggest New AI Features?

TL;DR

Google’s major AI rollout adds an AI Mode to Search, stronger Gemini 2.5 models, real-time Gemini Live interactions, and new creative tools for images and video. Veo 3 can generate videos with speaking characters, while NotebookLM gains a mobile app and Google AI Ultra bundles advanced features, YouTube Premium, and 30 terabytes of storage for $250 per month after an introductory discount.

Transcript

Google just announced and released some of the biggest updates in their AI lineup ever. And on the very top of the list, there is a brand new way to use Google.com to search. This is called AI mode inside of Google.com. And this is getting rolled out right now in the US as I speak. I've had it for about three months now as an experimental program. ... Read More

Key Insights

• Google Search AI Mode is a conversational search experience built directly into Google.com. It generates an answer with supporting links, suggestions, maps when relevant, and follow-up prompts, while the All tab continues to provide the traditional search results page.
• Gemini 2.5 Pro and Gemini 2.5 Flash are improved language models available in preview through Google AI Studio. The transcript identifies them as the top two models on the referenced LMArena chart, ahead of two ChatGPT models listed in third and fourth place.
• Gemini Live is designed for real-time interaction through a microphone, webcam, or shared screen. Inside Google AI Studio, it can observe what a user is viewing and provide immediate feedback, creating an experience compared in the transcript to having a tutor looking over the user’s shoulder.
• Gemini Diffusion is an experimental text model that applies a diffusion-style process to language generation. Unlike conventional language models described as predicting successive words, it visibly refines output toward an answer, as demonstrated through a fast mathematical example that arrived at 39.
• NotebookLM is available as a mobile app for iPhone and Android according to the description. The app lets users access AI-generated overviews and listen to audio discussions created from their own source materials while traveling, driving, or otherwise away from a computer.
• Veo 3 is Google’s state-of-the-art video generation model and can produce footage featuring characters who speak. The demonstrated clips included generated voices and accents, and access was described as part of the Google AI Ultra subscription rather than the free or standard Pro tier.
• Google AI Ultra is priced at $250 per month after an initial three months at half price. The plan includes Veo 3, Flow, higher access limits for Gemini 2.5 Pro and Deep Think, YouTube Premium, and 30 terabytes of storage.
• Whisk is a creative interface that combines a subject, scene, and style to generate images, with support for both text and image prompts. Users can refine results through follow-up instructions or animate them with Veo 2 while preserving visual consistency between frames.

Questions & Answers

Q: How does Google Search AI Mode work?

Google Search AI Mode adds a conversational experience directly to Google.com. Instead of primarily returning advertisements and blue links, it creates a synthesized response that can include supporting links, recommendations, a map, and suggested next steps. Users can ask follow-up questions from the same interface. The All tab remains available for returning to the traditional Google Search results page.

Q: Where can users try Gemini 2.5 Pro and Flash?

Gemini 2.5 Pro and Gemini 2.5 Flash can be selected in the Chat section of Google AI Studio at aistudio.google.com, where both were presented as preview models. Gemini 2.5 Pro was also shown as an available preview option at gemini.google.com. The studio provides additional testing options, including streaming access to Gemini Live for interactive voice and visual assistance.

Q: What can Gemini Live do with a camera and screen share?

Gemini Live can receive access to a user’s microphone, webcam, and shared screen, allowing it to respond to the situation in real time. The transcript describes it as working like a tutor looking over the user’s shoulder. A related Project Astra demonstration showed assistance with a bicycle repair by consulting a manual, finding a YouTube tutorial, checking email, and identifying relevant parts.

Q: What is Gemini 2.5 Pro Deep Think?

Gemini 2.5 Pro Deep Think is a new mode intended to push the model’s performance through enhanced reasoning and longer thinking. The transcript presents it as a separate mode associated with Gemini 2.5 Pro rather than as a standalone product. Higher access limits for Gemini 2.5 Pro and Deep Think are included among the benefits described for the Google AI Ultra subscription.

Q: How is Gemini Diffusion different from a language model?

Gemini Diffusion is described as an experimental model that applies a diffusion-like generation process to text. Conventional large language models are characterized in the transcript as predicting which word follows another based on training. Diffusion systems instead begin from a noisy or incomplete state and progressively refine it. Google demonstrated this approach by showing the model rapidly developing a mathematical solution that reached 39.

Q: What new capabilities does Veo 3 provide?

Veo 3 generates artificial video footage and adds the ability for generated characters to speak, including with different accents. The introductory footage shown during the presentation was described as having been created with the model. Veo 3 was available at the time covered by the transcript, but access was tied to the Google AI Ultra subscription, which costs $250 per month after the introductory discount.

Q: What is included with Google AI Ultra?

Google AI Ultra includes Veo 3, the Flow filmmaking tool, higher limits for Gemini 2.5 Pro and Deep Think, Whisk capabilities, YouTube Premium, and 30 terabytes of storage. It costs $250 per month after the first three months at half price. By comparison, the regular $20 Google AI Pro plan includes two terabytes of storage, while a free Google account includes 15 gigabytes.

Q: How does Whisk create and animate visual content?

Whisk lets users combine a subject, scene, and visual style through a simplified interface, then generate an image from text or image prompts. The demonstrated example created a miniature city inside a giant acorn. Users can refine an output with follow-up text instructions or animate it with Veo 2. In the example, a slow rotation preserved the scene consistently across the generated frames.

Summary & Key Takeaways

• Google Search is gaining an AI Mode that provides synthesized responses, source links, maps, suggestions, and conversational follow-up prompts. Users can switch back to the familiar results page through the All tab. The feature was rolling out in the United States when the presentation was recorded, following an earlier experimental period.

• Google expanded Gemini with improved 2.5 Pro and 2.5 Flash models, a Deep Think reasoning mode, an experimental text diffusion model, and enhanced live interactions. Google AI Studio offers chat access and a streaming experience that can use a microphone, webcam, and shared screen to provide real-time assistance or feedback.

• Google’s creative releases include Veo 3 video generation with speaking characters, the Flow filmmaking tool, Imagen 4, and Whisk for combining subjects, scenes, and styles. The Google AI Ultra subscription provides Veo 3, higher Gemini limits, 30 terabytes of storage, YouTube Premium, and other features for $250 per month after its introductory period.