The new Utopai X video generation model is built around the production workflows of a professional film studio.

Utopai Studios pairs its Utopai X video model with PAI, a platform that connects AI-assisted screenwriting, characters, shot generation, and editing.

Generating a video shot is not enough to make a film. You still need to retrieve a character's references, respect the established visual universe, maintain consistency across different takes, and integrate the selected images into a coherent sequence. Utopai Studios is developing this approach with two complementary tools: Utopai X, a video generation model, and PAI, a production platform that preserves the information necessary to create an audiovisual project. The former reached second place in Artificial Analysis's Text-to-Video with Audio ranking, with 1,150 Elo points in the September 29, 2026 results. The latter organizes the work around the script, characters, sets, shots, and their various versions. A video model developed alongside actual productions

Utopai Studios does not approach artificial intelligence merely as a supplier of generative models. The company also finances and produces films and series, allowing it to test its tools against the real-world constraints of long-term audiovisual projects. Technical teams thus work directly from production materials: scripts, visual research, approved designs, shooting schedules, and successive iterations of images.

This organization influences the training of Utopai X. Directors, cinematographers, and artists evaluate generated sequences, select specific takes, and explain why others are unsuitable. These choices provide additional data for the research teams, who can then address problems identified during the filmmaking process.

Utopai X relies on a pre-existing video model adapted to the company's needs. The studio particularly highlights its ability to render reflections, caustics—light patterns produced by reflection or refraction—and the spatial coherence of objects within a shot. These are highly sought-after qualities in scenes featuring complex interactions between light, surfaces, and camera movements.

These observations stem from the studio's internal creative evaluations. They describe areas where its teams feel they achieve strong results, though they do not constitute independent measurements of character continuity or visual stability across multiple shots. A second-place finish on Artificial Analysis, on a specific date

The visual quality of Utopai X underwent external evaluation on the Artificial Analysis video leaderboard. In the results from September 29, 2026, the model held the second position globally in the Text-to-Video with Audio category, with 1,150 Elo points, behind Wan 3.0. It was also the highest-ranked model among those developed by U.S.-based companies in this category.

Artificial Analysis uses blind side-by-side comparisons between two videos generated from the exact same prompt. Evaluators choose their preferred sequence without knowing which model generated it. The votes are then used to calculate Elo scores and establish the leaderboard ranking.

While this protocol provides an indication of evaluators' visual preferences, it does not directly test the full range of constraints in a feature film production. The ability to maintain an identical character across multiple scenes, precisely reproduce a requested edit, or preserve match cuts between two shots requires additional evaluation.

Furthermore, the reported score represents a snapshot of the leaderboard as of September 29, rather than a permanent position. Results are subject to change as new models are introduced and additional votes accumulate. PAI preserves the script and production decisions

The operation of Utopai PAI, short for Production Assistive Intelligence, is built around a complete audiovisual project rather than a series of independent generations. The platform analyzes the script to identify scenes, characters, locations, and props, then assists in building a shot list and preparing the corresponding visual assets.

The Production Library gathers the film's characters, objects, environments, and reference materials. Different versions remain tied to the project, making it easy to retrieve a previously approved element when preparing a new shot. The goal is to prevent each generation from requiring a complete re-prompting of the visual world, character traits, or directorial intent.

PAI also features a visual workspace organized around a canvas. Teams can arrange their references, develop scenes, and group elements related to a sequence within nested spaces. The project's structure remains accessible throughout the exploration, generation, and revision phases.

AI assistants, called Production Assistants, can leverage project information to prepare shots, suggest variations, and help execute specific tasks. However, directorial choices, output selections, and final approvals remain strictly in the hands of the human filmmakers. Multiple takes and a timeline over isolated video clips

When a shot is generated with Utopai X, it remains anchored to its place in the project. Teams can produce different versions, revert to a previous take, and compare results before selecting which ones to keep.

The platform integrates a timeline designed to assemble chosen sequences and review their pacing and flow. This feature allows filmmakers to evaluate a video within the context of its edit, rather than judging its quality purely as a standalone clip. PAI also supports the creation of images, videos, and audio content across several models integrated into its environment.

Exports then allow work to continue seamlessly in Adobe Premiere Pro or DaVinci Resolve. Utopai also mentions the ability to export videos in 4K. However, the Utopai X model should not be mistaken for a natively 4K video generator: the export resolution offered and the native resolution used to generate a sequence are two distinct technical specifications.

For professional teams, PAI also includes collaborative features. Comments and approvals are tied to specific version numbers, while a dashboard allows project milestones to be tracked. Traceability features preserve the history of