PixVerse (Aishi Technology)

What PixVerse is

PixVerse is an AI video generation company founded in 2023, with teams distributed across Asia and the United States ^[1]. It describes itself as a global AI video generation platform, and says its models are developed entirely in-house rather than assembled on top of someone else's foundation model ^[1]. In Chinese financial reporting the company appears as Aishi Technology ^[3].

"AI video generation" here means producing a video clip from a written prompt (text-to-video) or from a still image that is then animated (image-to-video). Both are computationally heavier than image generation, because the model has to keep objects, motion and lighting consistent across many frames rather than a single picture.

Origins and leadership

The co-founder and chief executive is Changhu Wang, and the co-founder and president is Jaden Xie ^[1]. Chinese financial press, covering the company's funding round, names Wang Changhu as founder and chief executive of Aishi Technology and reports the same round as 2.98 billion yuan, with the C+ tranche led by Alibaba ^[3].

The models

As of 28 February 2026, the company's V5.6 model ranked second worldwide in both image-to-video and text-to-video on Artificial Analysis, a site that compares generative models on public benchmarks ^[2]. V5.6 is built on the company's own diffusion transformer and generates 1080p video in under a minute with native audio ^[2]. A diffusion transformer is a architecture that combines two ideas: diffusion, which learns to turn noise into a clean sample step by step, and the transformer, the attention-based network behind most current language and image models. "Native audio" means the sound is produced by the same generation process as the picture, instead of being added afterwards.

In January 2026 the company launched R1, which it calls the world's first real-time world model ^[1][2]. A world model, in this context, is a model that learns how an environment behaves and can simulate it, so that a user can act inside the simulated scene and see it respond immediately rather than waiting for a clip to be rendered.

Products and developer tools

The product line covers a web app, a command-line tool and plugin, Canvas, an agent, and an API platform ^[1]. The model menu is not limited to the company's own systems: it lists its own V6 and R2 alongside third-party models such as Seedance 2.5, MiniMax H3 and Flux 3 ^[1]. Offering outside models next to in-house ones lets users pick a model per job without leaving the platform. The GitHub organisation publishes an official MCP server, a command-line tool and an agent skill library, aimed at developers who want to call the models from their own software or from AI agents ^[4].

Funding and scale

PixVerse closed its Series C on 12 March 2026, led by CDH Investments, and said the round made it a unicorn ^[2]. The company called it the largest funding round in the AI video generation category in Asia ^[2]. It opened a global office in Singapore to handle enterprise sales in North America and Asia ^[2].

By September 2026 the company claimed more than 150 million users in 177 countries and regions ^[1], and reported more than 2.1 billion videos generated and 16 million monthly active users at that point ^[2].

Where it sits

PixVerse is one of a small group of video-generation companies that build their own models and also run a consumer-facing product. Its public positioning rests on three things: benchmark placement for its own models ^[2], a stated policy of in-house development ^[1], and a product surface that reaches from a browser app to APIs and agent tooling ^[1][4]. The funding round and the user figures describe how far that combination has been carried; the benchmark rankings and the R1 claim are the company's own statements about where its technology stands.