Google, OpenAI tighten daily limits on AI image and video generation

Google and OpenAI have introduced new daily caps on free image and video generation as soaring global usage strains computing resources. The restrictions aim to balance accessibility, manage GPU load, and maintain service quality for millions of users.

Google and OpenAI have imposed fresh limits on their AI image and video generation tools as soaring global demand overwhelms available computing capacity. The rapid rise in usage—driven by excitement around Google’s Nano Banana Pro image model and OpenAI’s Sora video generator—has prompted both companies to revise their daily free-access policies.

OpenAI has capped Sora's free video generation at six videos per day, though paying customers under ChatGPT Plus and Pro plans will retain unrestricted access. Users can also purchase additional video generations if required.

Bill Peebles, Head of Sora at OpenAI, confirmed the update on X, noting the strain on infrastructure. He admitted that the system’s GPUs were “running hot,” adding that the change would help more users access the technology without disruption.

Google restricts free image creation on Nano Banana Pro

Google has also tightened limits on image generation through its Nano Banana Pro model, restricting free users to two image generations or edits per day. The model, powered by Gemini 3 Pro Image, was launched with significant anticipation, leading to unusually high traffic shortly after release.

The company has introduced new variable limits for the broader Gemini 3 Pro tier as well. Google’s support page now warns that “daily limits may change frequently” for free users accessing Gemini 3 Pro Thinking capabilities. In contrast, Google continues to allow up to 100 daily images on the earlier Nano Banana version, which is powered by the less demanding Gemini 2.5 architecture.

Compute pressure highlights AI’s growing resource demands

The temporary tightening of quotas coincides with a holiday weekend in the United States—typically a period of heavy usage as users experiment with creative tools. Both companies noted that the restrictions underscore the significant computational and energy resources required to generate high-quality images and videos.

OpenAI CEO Sam Altman recently clarified comments about “melting GPUs,” explaining that while processing units can run intensely hot under heavy loads, modern data centres use sophisticated cooling systems to prevent actual damage.

The new limits, while temporary for some users, reflect the ongoing challenge AI companies face as they scale increasingly complex generative systems for global audiences.