Summary of Google I/O 2025 product showcase - Stitch, Project Astra, Veo3, Imagen4, Flow,Joules and more..
Google I/O 2025 delivered serious upgrades—Project Astra now sees and remembers, Stitch turns ideas into UI, Veo adds sound to AI video, and Search gets a Copilot. I handpicked the releases I loved
Google shipped tools that aren’t just demos—they're useful, usable, and pushing boundaries across design, search, video, and productivity.
Some tools feel like future prototypes (Project Astra, Flow), while others are quietly changing how we work every day (Stitch, Gemini in Chrome, AI Search Mode). In this post, I break down what stood out to me—not just as an observer, but as someone who builds and experiments with these tools constantly.
No Time to Read? Here's the Scoop
Project Astra : Real-time AI that sees what you see and hears what you hear—Google's most impressive demo this year. This is what a universal AI assistant could look like.
Stitch : Conversational UI design assistant that builds mobile/web screens step-by-step and exports directly to Figma. Still in Labs but shows crazy potential. Loved this one.
Veo 3 + Flow : Google’s image and video model now brings native audio—think dialogue and background sounds directly into the clip. No one else is doing this yet.
Gemini in Chrome Finally, a browser assistant. Gemini now lives right inside Chrome and can help you work with the content on any website.
AI Search Mode Search gets an AI overhaul with a new tab that handles longer, more complex queries. SEO will be difficult
Google I/O 2025: My Favorite Highlights
I watched the Google I/O 2025 keynote with great anticipation and they have delivered. From practical upgrades to wild demos, here’s what stood out for me.
1. Project Astra – A Glimpse at the Future of AI Assistants
This was my favourite moment. In the demo, Astra literally sees through your phone's camera, identifies your surroundings, and responds live.
You can ask it, "What's this?" and point to anything. It remembers what it saw earlier too—like tracking where you left your glasses. That’s wild.
The screen-sharing ability combined with contextual memory is exactly the direction assistant tech needs to go. Google is clearly thinking beyond chatbots and heading toward real-world companions.
Watch the demo:
2. Stitch – Conversational UI Design in Figma’s Shadow
Still in Labs, but Stitch might just be the missing piece for founders, PMs, and indie hackers who want to go from idea to UI without opening Figma.
You describe the app idea, say whether it’s for mobile or web, and Stitch builds the screens for you step-by-step. Then you can copy the designs you liked and paste it in Figma. This feels like the design tool for non-designers—finally.
Try it at:
I tried making a UI with Stitch and it’s quite fascinating.
3. Veo 3 + Flow – The Audio-First Take on Video Generation
I didn’t expect to be this impressed, but Veo 3 now brings native audio directly into video generations—environmental sounds, footsteps, dialogue—it’s baked into the generation pipeline.
This powers Flow, an early-access tool for AI filmmakers to create longer, coherent cinematic scenes. Google’s positioning Flow for serious creators, and it shows.
Watch Flow's capabilities:
4. Gemini in Chrome – Actually Useful Browser AI
This one slipped in without much fanfare, but I found it super useful. Gemini now runs inside Chrome and can help you analyze or summarize any webpage. It’s like having a researcher on tap.
Search History with AI
Shop with AI
Create your own Chrome Theme with AI
5. Search AI Mode – The Biggest Redesign in a Decade
Search now has an AI Mode tab. It can handle multi-part queries and longer, more abstract questions. Google says early testers are writing 2–3x longer queries compared to regular search.
It’s subtle, but this is one of the biggest shifts to search behavior I’ve seen. It almost feels like a Copilot for curiosity.
Other Great Releases Worth Noting
Gemini 2.5 Pro: Most advanced LLM from Google with Deep Think and better reasoning. SOTA across most benchmarks. Read more
Imagen 4 : Photorealistic Image Generation, Nailed . Read More
Google Diffusion Model : Google’s new diffusion model is built for fast generation—really fast. Token generation speed is among the fastest - 5 (five) times the speed of Gemini 2.5 Pro for coding tasks. It achieves a sampling speed (excluding overhead) of 1479 tokens per sec
Project Mariner: Google's take on agentic AI, helping users complete multistep tasks like job search and grocery planning. Read More
Virtual Try-On: Enhanced visual preview for clothing—feels like AR meets Pinterest. Read More
Google Beam: Turns 2D calls into 3D video calls using multi-camera setups. HP ships the first devices later this year. Read More
Real-Time Speech Translation in Google Meet: Talk in your language, get live translated responses. Read More
Joules: AI for developers. Asynchronous coding agent. Read More
Firebase Upgrades: Better tooling for app dev, debugging, and scaling. Read More
Check out my previous articles:
