Google Introduces Veo 3 With Native Audio Generation; Deep Think Mode Brought to Gemini 2.5 Pro; More Updates From Google I/O 2025 Keynote

The much-awaited Google I/O 2025 annual developer conference has commenced and the already-held keynote address has unveiled a lot of the brand’s innovations from the past few months – of which most of them are AI-related. The brand has now introduced native audio generation capability on its Veo 3 video generation model, while a Deep Think reasoning model has been brought to Gemini 2.5 Pro as well.

Check out more about these updates and regarding the other updates announced at the I/O 2025 Keynote below.

Veo 3 With Native Audio Generation

The introduction of the Veo 3 model is one among the notable announcements that has been made by the brand at the I/O 2025 Keynote. With Veo 3, in addition to the improved video generation quality when compared to the previous Veo 2 model, users from now on will be able to add dialogues between the characters, as well as add sound effects and background noise. Thus as pointed out by the brand, Veo 3 marks the end of the silent video generation era.

As of now, the Veo 3 model is available on the Gemini app, and Google AI Ultra subscribers in the United States will be able to enjoy this new video generation capability. Given below is a video generated using the latest Veo 3 model.

Source: 1

Deep Think Model on Gemini 2.5 Pro

On the Gemini 2.5 Pro, Google has brought its new and enhanced reasoning mode – Deep Think. It has been made based on the brand’s research in parallel thinking techniques and it will explore multiple hypotheses first before it actually responds to the users’ requests.

Deep Think thus handles complex math and coding problems much more effectively and in addition to this, Gemini 2.5 Pro’s Deep Think was able to score greatly on the USAMO 2025 benchmark – one among the hardest math benchmarks. On the LiveCodeBench too it has taken the lead, scoring 84% on MMMU.

As of now, in order to gain feedback, Google has made Deep Think available to certain safety experts, and over the coming weeks the brand plans to share it with trusted testers too via the Gemini API

Source: 2

Other Updates From I/O 2025 Keynote

Apart from these, at the I/O 2025 Keynote, other announcements were made too. Updates to the Gemini 2.5 Flash and Pro models have brought native audio outputs, and along with the Veo 3 video generation model, the Imagen 4 image generation model was also introduced. Updates have also been brought to Deep Research and Canvas as well.

For an AI-based shopping experience, Shopping in AI Mode, Virtual Try-on Tool, and Agentic Checkout have been unveiled. In Google services too, AI related updates/integration have been brought. In Gmail – Personalized Smart Replies, in Meet – Speech Translation, and in Chrome – Gemini integration are certain notable announcements.

The support for Anthropic’s Model Context Protocol (MCP), Synth ID Detector, expanded access on AI Overviews, Search Live in AI Mode and Lens, and expanded access on Gemini Live’s camera and screen sharing features are other announcements. The brand also introduced Beam – a 3D AI-first video communication platform, Jules – an asynchronous AI coding agent, and Flow – a new AI flim-making tool. An early demo of Google’s Android XR on smart glasses was also shared. Finally, AI Pro and AI Ultra new subscription plans were also announced, priced respectively at USD 19.99 per month and USD 249 per month.