Alibaba-backed PixVerse launches real-time AI video tool

BEIJING — An Alibaba-backed startup is pushing artificial intelligence toward real-time, interactive video creation.

PixVerse on Tuesday released an AI tool that allows users to control how a video unfolds while it is being generated. And just like a movie director, users can instruct characters to cry, dance or pose, with actions occurring instantly as the video continues.

Real-time AI video generation can create “new business models,” co-founder Jaden Xie told CNBC in an interview translated from Chinese. The possibilities, he said, include a world where users could influence how a micro-drama unfolds, or play an “infinite” video game that’s not constrained by pre-designed storylines.

Founded in 2023, PixVerse raised more than $60 million in the fall, with Alibaba leading the round and participation from Antler.

Xie said the company is close to completing another funding round, without disclosing the amount. More than half of the participating investors are from overseas, he said.

PixVerse’s latest AI tool highlights how China-based teams are giving their competitors a run for their money in AI-generated video tools.

Except for Israeli startup Lightricks, the top eight AI video generation models tracked by AI benchmarking firm Artificial Analysis are Chinese companies. Many offer faster generation speeds and far lower usage fees than OpenAI’s advanced, premium AI video model, Sora 2 Pro.

OpenAI’s Sora first drew global attention nearly two years ago when it showed off a new text-to-video generation model, but it was not publicly available until December 2024. By then, several Chinese teams had already released competing tools to users globally.

“Sora still defines the quality ceiling in video generation, but it is constrained by generation time and API cost,” said Wei Sun, a principal analyst at market research firm Counterpoint, referring to the way users are charged for access to use AI models.

“Chinese players are taking a different path. They are turning video generation into a scalable, low-cost, high-throughput production tool.”

Just last month, Beijing-based startup Shengshu said its TurboDiffusion video framework, developed with researchers from Tsinghua University, could create videos 100 to 200 times faster with minimal quality loss.

PixVerse’s latest AI tool has removed most of the waiting time — and aims to be more than just a special effect. Its tools are embedded within the company’s social media-style sharing platform, which surpassed 16 million monthly active users in October.

Real-time video generation eliminates the gap between content creation and distribution, Xie said, reshaping how users interact with AI-generated content.

Xie aims to reach 200 million registered users in the first half of this year, up from 100 million in August, and roughly double the team to nearly 200 employees by year’s end.

PixVerse primarily serves users outside China through a web-browser interface and smartphone app.

Compared to China-made AI video tools, “most of the U.S. products are relatively simplistic and minimal” in user interface and experience, said Alyssa Lee, chief of staff at DataHub and a former vice president at Bessemer Venture Partners.

Lee said scenario-specific AI video tools offer clearer monetization paths and pointed to Adobe as a traditional software provider facing pressure. Adobe, long a standard for video and design software, has seen its stock stagnate in recent months, she said, suggesting its “all-in-one creative suite is vulnerable to being unbundled by all these creative AI marketing tools.”

PixVerse estimated annual recurring revenue of $40 million in October.

It’s not the only one generating revenue.

Kling, an AI video tool developed by TikTok rival Kuaishou, posted nearly $100 million in revenue in the first three quarters of 2025, based on CNBC calculations from public disclosures.

For now, Xie said that PixVerse is prioritizing technology development over commercialization and claimed the company has sufficient funding to operate for a decade.

Addressing concerns about low-quality AI-generated content, often referred to as “slop,” Xie compared the current stage of development to the early years of computer graphics, arguing that quality improves as technology matures.

“At the beginning, there will be good and bad [content], but gradually the fittest will surely survive … and then some people will improve the technology, and truly meet human needs for emotional and spiritual value.”