AI Film Studio Utopai's Video Model Ranks Second on Text-to-Video Leaderboard

Artificial Analysis (AA) has published its latest text-to-video leaderboard, and the surprise is not the winner, which is Alibaba's Wan 3.0, but the runner-up. Second place went to Utopai X, a model from Utopai Studios, a film company rather than a big-tech platform or a model lab, with 1,150 points, just seven behind the leader.

The ranking is based on blind voting by reviewers recruited by AA, who watch two videos generated from the same prompt without knowing which model made them and pick the one they prefer. QbitAI highlighted two of the tests, a handheld tracking shot of someone pushing a wooden boat into shallow water and a plush hyena under a lamp that brightens from dark to full, to show how the model handles physics, camera movement, and lighting.

According to QbitAI, Utopai Studios has been called the world's largest independent AI-native film company by Variety, and Forbes estimates its valuation at roughly $1 billion. Its CEO and co-founder, Cecilia Shen, previously worked at Google X and was named to Forbes' AI 30 Under 30 this year.

The model debuted alongside a new production platform called PAI, with the model embedded inside it. A director drops in a script and chooses a style; the system then breaks the script into scenes and shots and lays out characters, sets, and multiple versions of each shot on a single canvas, carrying a project from script through to editing.

One unusual PAI feature is an impact check: if a director tries to insert a new shot between two existing ones, an agent warns that the following two shots will no longer match in terms of where characters are looking, and flags them for regeneration, while the director still makes the final call. The workflow has been used on real projects, including an animated short starring NBA player James Harden and a September collaboration with DeNA, the developer behind several Pokemon games, to explore AI-assisted animation.

On the technical side, Utopai X is built by post-training on MiniMax's open-source video model H3. QbitAI notes that AA placed Utopai X and MiniMax H3 in the same blind test, effectively creating a controlled comparison: Utopai X came closer to the top tier than its base in seven of ten capabilities, with the biggest gains in audio synchronization, camera control, and physics. In the live-action film category it finished just four points off the top score, and it placed first for social media content and architecture and real estate.

Utopai attributes that film-literate preference to the studio floor. Its technical team and film projects sit inside the same company, so the problems artists hit while designing characters, building environments, and organizing shots flow back into the models, agents, and workflows along with production data and expert feedback. Internally, an Elo rating system combines human preference evaluations with automatic votes from vision-language models.

The film slate includes Cortes, written by Oscar-nominated screenwriter Nicholas Kazan and set in 1519 during the Spanish entry into the Aztec Empire, targeting a wide theatrical release in 2027; the sci-fi series Space Nation, also planned for 2027; the family animated feature The Most Serious Fart, based on Mike Bender's children's book with Shrek writer Joe Stillman as co-executive producer; and the German-Korean live-action film Half Moon, directed by Hyo-joo Yang, who won a Silver Bear at Berlin for her short Broken Night.

Commercially, QbitAI cites Forbes reporting that Utopai has pre-sold some international distribution rights to Brazil's Globoplay and a distribution arm of a major German broadcaster, at prevailing market prices for non-AI content. If delivered as planned, the Space Nation and Cortes projects alone could bring in up to $110 million, and more than twenty projects are in various stages of development across theatrical, TV, streaming, and digital platforms.