Google Bets Big on AI Video for the Masses While OpenAI Quietly Retreats From Sora

Google is making its most aggressive move yet to embed AI-generated video into the daily lives of ordinary consumers — and it’s doing so at precisely the moment its chief rival is pulling back from the same ambition.

At its annual Google I/O developer conference this year, the company unveiled a sweeping expansion of its AI video capabilities, integrating its Veo model directly into products that hundreds of millions of people already use. Google Photos. Google Workspace. The Android operating system itself. The message was unmistakable: AI video generation isn’t a novelty act anymore. It’s infrastructure.

As TechRadar reported, Google is now positioning AI video not as a standalone creative tool for professionals and hobbyists, but as a feature woven into the fabric of products people use without thinking. That’s a fundamentally different strategy from the one OpenAI has pursued with Sora, its own video generation model — and the divergence between the two companies tells a revealing story about where the AI industry is headed.

OpenAI launched Sora with enormous fanfare in early 2024, releasing demo videos that stunned the internet with their photorealism. A woman walking through a neon-lit Tokyo street. Woolly mammoths trudging through snow. The clips were visually arresting, and they sparked immediate conversations about the future of filmmaking, advertising, and visual media. But the product’s actual rollout has been anything but smooth.

Sora was initially made available to a limited group of users in December 2024, bundled with ChatGPT Plus and Pro subscriptions. Demand overwhelmed capacity almost immediately. OpenAI had to pause new signups within hours. And when users did get access, many found the tool’s outputs inconsistent — impressive in short bursts, but prone to the kind of visual artifacts and logical failures that make AI-generated video unreliable for professional use. Hands with too many fingers. Physics-defying object movements. Characters whose faces shift mid-scene.

Then came the cost problem. Generating even short video clips with Sora consumed significant computational resources, and OpenAI struggled to offer the tool at a price point that made sense for casual users while still covering its infrastructure costs. According to TechRadar, OpenAI has effectively pulled Sora back from broad consumer availability, limiting access and scaling down its ambitions for the product — at least for now.

Google, meanwhile, is taking the opposite approach. Rather than building a standalone video generation product and hoping users come to it, Google is embedding Veo’s capabilities inside tools people already depend on. The integration into Google Photos is particularly telling. Users will be able to generate short video clips from their own photo libraries, turning static memories into animated sequences with minimal effort. No prompting expertise required. No separate app to download.

That’s a significant philosophical difference.

OpenAI built Sora as a destination. Google is building Veo as a utility. And if the history of consumer technology teaches anything, it’s that utilities win. Features that reduce friction and meet people where they already are tend to achieve mass adoption far faster than standalone products that require users to change their behavior.

Google’s approach also reflects a hard-learned lesson from the company’s own history with AI. When it launched Bard — later rebranded as Gemini — the chatbot initially struggled to compete with ChatGPT in part because it existed as a separate product rather than being deeply integrated into Google’s existing services. Google has since corrected course aggressively, embedding Gemini into Search, Gmail, Docs, and virtually every other product in its portfolio. The Veo integration follows the same playbook.

The technical underpinnings matter here too. Google’s Veo model, now in its third generation, has shown marked improvements in temporal consistency — the ability to maintain coherent objects, lighting, and physics across multiple frames. This has been one of the hardest problems in AI video generation, and while no model has fully solved it, Google’s progress has been notable. Veo 3, announced at I/O 2025, can generate clips with synchronized audio, a capability that few competitors have demonstrated at comparable quality levels.

But let’s not overstate the case. AI-generated video remains deeply imperfect. Even Google’s best outputs occasionally produce the uncanny visual glitches that have become a hallmark of the technology. And there are serious unresolved questions about copyright, consent, and the potential for misuse — questions that become exponentially more urgent when you’re putting these tools in the hands of billions of users through products like Google Photos and Android.

The copyright issue alone could reshape this entire market. AI video models are trained on vast datasets of existing video content, much of it created by filmmakers, photographers, and artists who never consented to having their work used as training data. Multiple lawsuits are working their way through courts in the United States and Europe. The outcomes could impose significant constraints on how companies like Google and OpenAI deploy these tools.

Google has attempted to get ahead of this by implementing content credentials and watermarking systems for AI-generated video, using its SynthID technology to embed invisible markers in generated content. Whether these measures will satisfy regulators, rights holders, and the public remains an open question.

OpenAI, for its part, hasn’t abandoned Sora entirely. The company continues to develop the model and has signaled that it plans to reintroduce it more broadly once it can solve the scaling and cost challenges that hampered the initial launch. But the window of opportunity may be narrowing. Every month that Google spends embedding AI video into its consumer products is a month that OpenAI loses in the race for mainstream adoption.

And Google isn’t the only competitor. Meta has been developing its own video generation models, and startups like Runway, Pika, and Kling have carved out significant niches in the professional and prosumer markets. The field is crowded and moving fast. OpenAI’s first-mover advantage with Sora — which seemed so commanding just a year ago — has eroded considerably.

There’s a broader strategic dimension to this as well. For Google, AI video generation is part of a larger effort to make its AI capabilities the default layer underlying all digital activity. Search, communication, productivity, creativity — Google wants Gemini and its associated models to be the invisible engine powering all of it. Veo’s integration into consumer products isn’t just about video. It’s about establishing AI as a Google-controlled utility that users interact with constantly, often without realizing it.

For OpenAI, the challenge is different. The company has built its business primarily around ChatGPT, a single product that users actively choose to engage with. Expanding beyond that into video, voice, and other modalities requires either building entirely new products — expensive and risky — or partnering with platform owners who control distribution. OpenAI’s partnership with Microsoft gives it access to some distribution channels, but nothing approaching the scale of Google’s consumer footprint.

So the competitive dynamics are shifting. Not because one company’s technology is dramatically superior to the other’s, but because distribution and integration strategy are proving to be at least as important as raw model quality. Google understands this instinctively. It’s a company that has spent two decades perfecting the art of embedding services into the daily habits of billions of people. OpenAI, for all its technical brilliance, is still learning how to do that.

The next twelve months will be decisive. If Google can deliver a reliable, useful AI video experience inside products like Photos and Workspace — one that feels natural rather than gimmicky — it will have established a lead that’s very difficult to overcome. If the technology disappoints, or if regulatory action constrains deployment, the race stays open.

Either way, the contrast between Google’s push forward and OpenAI’s tactical retreat marks a turning point. The era of AI video as a dazzling demo is ending. The era of AI video as a mundane, embedded feature of everyday software is beginning. And right now, Google is better positioned for that transition than anyone else.