Brief Summary
This video provides a first look at the massive AI updates announced at the Google IO conference, highlighting key features, product upgrades, and changes across Google's ecosystem. The presenter discusses the new Google AI plans, including the Pro and Ultra plans, and showcases impressive updates to Gemini Studio, including Deep Research and Canvas improvements. The video also covers the new Flow product for video generation, live translation in Google Meets, AI-enhanced Google Search, AI Try On for shopping, and Google Labs experiments like Project Mariner.
- New Google AI plans: Pro and Ultra
- Gemini Studio updates: Deep Research and Canvas
- Flow product: video generation with audio
- AI-enhanced Google Search with a new AI mode
- Project Mariner: AI agent for computer operation
Overview
Google's IO conference featured a significant batch of AI releases, including updates, features, and products across their ecosystem. Google Search now has a new tab with an AI mode. The video provides a first look at the most significant parts of these releases, with the presenter having upgraded to the new $250 plan to try out the available features. The presenter emphasizes the incredible nature of the AI updates and the extensive changes across Google's products.
Pricing
Google has introduced new AI plans for their Gemini products and other services. The original Gemini plan at $20 a month has been transformed into the Google AI Pro plan, and a new Google AI Ultra Plan is available for $250 a month. Accessing the Ultra Plan can be challenging, especially in Europe, often requiring a VPN. The Ultra Plan provides access to a wide array of new features and updates across Google's product suite, though some are still rolling out. Many features are freely available, while others are exclusive to the paid plans.
Gemini Models
The presenter focuses on two standout features: the new Flow product for video generation with audio and the updates to Gemini Studio. Gemini Studio includes V3 generation and image and 4 generation. The presenter highlights the improvements to Deep Research and Canvas, as well as updates to mobile apps on Android and iOS that include a ChatGPT-like voice assistant. The new Gemini 2.5 Pro Deep Think model outperforms existing benchmarks but is not yet available. A Flash model, accessible today, closely matches top models at a lower cost.
Deep Research & Canvas
The presenter compares Gemini's Deep Research with ChatGPT, noting that Gemini's results were more specific to the query about a recovery plan for tendonitis from kite surfing. The updated Deep Research now supports uploading documents like PDFs. Canvas, the interactive feature for displaying code and writing, has also received major updates. In a comparison, Gemini's Canvas created a more interactive and 3D solar system web app compared to ChatGPT, even after a follow-up prompt to ChatGPT with implementation details from Gemini.
Flow (Veo3)
The new VO3 model now includes audio generation. Examples include a video of a cat with a hat typing, featuring spot-on sound effects and ambience. The presenter showcases several other examples, highlighting the improved audio and video quality. The video model is a significant advancement, incorporating a scene builder that allows users to stitch multiple clips together and edit them within the interface. Users can also use text-to-video prompts, adjusting settings for the highest quality output. The Flow application allows users to mix and match reference images to generate unique videos, though this feature is not yet fully functional with the new models.
Live Translation & AI Search
Google Meets will soon feature live translation, starting with Spanish and English. Google Search has been modified with a new AI-enhanced search tab, providing a ChatGPT/Perplexity-like experience directly within Google Search. This new tab offers results without ads and adapts the interface based on the search query, potentially including analysis tools, Python code execution, and custom graph generation in the future. Users can enable this AI mode in the search labs settings.
AI Try On
The new Google AI search mode will include shopping features, such as an AI Try On that fits clothes onto a 3D model of the user's body. This feature allows users to try on clothes virtually through the web. The presenter appreciates Google's ability to ship production-ready versions of technologies that smaller startups have been developing.
Google Labs
Google Labs has added numerous new experiments, including Jules, a competitor to OpenAI's Codeex for developer products. Project Mariner, an AI agent that operates a computer using mouse and keyboard, is also available, featuring a "teach a task" function that allows users to train the agent on specific tasks. Many of these experiments are free to try at labs.google/experiments. Google AI Studio is their developer interface, unifying various AI tools and models. The new imaging model is expected to be available in AI Studio soon, along with the new Flash model. Other upcoming features include Gemini Chrome, which integrates an AI assistant into the Chrome browser, and updates to the imagen model for more realistic image generation with improved text rendering.
