Honest Google IO Review | Open AI Takes the W, but veo AI Video is dope!

Google’s AI keynote delivered a mixed bag: much of it felt incremental, but one announcement—vo, a high-quality 1080p video generator—stood out enough to draw direct comparisons with OpenAI’s Sora. The most concrete takeaway from the Google side is that vo can produce long, coherent, frame-consistent video from text prompts, including camera-like motion and the ability to add or animate objects within an existing scene. In practice, the demo clips looked impressively stable across frames, with detailed environments (neon cityscapes, futuristic sprawl, suburban streets) and convincing “real-world” shots such as close-ups of food and aerial-style views.

Still, the strongest moments came with caveats. The clearest examples of vo’s consistency were sometimes undermined by the way the demo footage was presented—compression artifacts made some clips look blurry or “mushy,” and at least one showcased transformation sequence appeared less crisp than expected. Even so, the underlying capabilities were treated as the real story: multi-step generation, object insertion/animation, and photo-to-video style effects via Google’s image generation pipeline (including an “alpacas” prompt). The presenter’s overall verdict was that vo is competitive, but not definitively ahead—Sora was viewed as slightly more impressive based on first impressions.

Beyond video generation, Google’s IO rollout pushed AI deeper into search and productivity. AI Overviews are rolling out in Google Search, and an “Ask photos” feature is planned for Google Photos. Workspace Labs is slated to improve email summarization and meeting summarization, while Google also previewed prototypes of AG agents positioned as “virtual teammates.” On the search front, Google highlighted multi-step reasoning for complex questions and planning tasks, plus video understanding inside search results—an expansion aimed at making search feel more like an assistant than a link list.

The transcript also frames a broader competitive tension: OpenAI’s earlier announcements were seen as overshadowing Google’s keynote, with many observers leaning toward OpenAI’s “spring updates” (including GPT-4o and a ChatGPT overhaul) as more exciting. Google’s response, in this telling, is a sprint to catch up—especially in agents, where demos included email organization into spreadsheets, replying to emails, and browser agents that can operate across multiple websites (such as updating addresses). The practical concern raised is that demos may not translate cleanly into real-world reliability.

Finally, the transcript notes a major personnel shift at OpenAI: Ilya Sutskever is leaving, with Jacob Devlin named as the new Chief Scientist. The segment ends by tying the competitive race to hardware acceleration as well—Grok Inc’s chip-driven demo showed AI models controlling a retro fighting game (Gemma 7B versus ChatGPT 4 Turbo), reinforcing the idea that speed and deployment matter as much as model quality.

Overall, Google’s keynote mattered most for vo and the direction of travel: more assistant-like search, more agentic workflows, and tighter integration across Photos, Workspace, and Android—while OpenAI’s lead is portrayed as still intact, particularly for video generation and assistant polish.