Wonder 3D from Autodesk: The AI 3D Model Generator That Has Already Changed the Game

A major development in the world of 3D graphics has simultaneously delighted and alarmed industry professionals. Autodesk, the well-known developer of modeling software (3ds Max, Maya, AutoCAD), has released Wonder 3D — a powerful generative AI capable of creating full-fledged 3D models from text descriptions or simple images. But the story of this technology begins in cinema: Wonder 3D is built into the Autodesk Flow Studio platform, which is the successor to Wonder Studio — a service founded by actor Ty Simpkins (“Iron Man 3”) and director Nikolas Todorovic. Originally, the startup allowed for the automatic replacement of live actors with 3D characters without expensive motion capture. After acquiring Wonder Dynamics in 2024, Autodesk added the ability to generate the models themselves from scratch.

Technically, Wonder 3D offers three key modes. First, Text-to-3D: you type “robot samurai in red armor,” and the AI generates a volumetric model. Second, Image-to-3D: any sketch, photo, or concept art is transformed into a 3D object. Third, a combined mode where the AI first creates several 2D concept variants and then builds a 3D model based on the chosen one. Finished models are exported in OBJ format and can be easily transferred to Maya, Blender, Unreal Engine, or sent to a 3D printer. Autodesk emphasizes that the models are fully editable — the AI creates a base that an artist can refine, rather than just generating a “baked” static image.

This is already accessible to users: the free tier offers 300 credits per month (that’s 15 generations, as each costs 20 credits), while paid subscriptions range from $6 to $51 per month on an annual contract. On the free tier, video export is limited to 720p, but that’s sufficient for testing 3D generation. However, the main controversy has unfolded not around price, but around ethics. French journalists from the industry publication 3DVF decided to test how Wonder 3D handles copyright filtering. In response to a direct request for “Super Mario,” the AI politely refused — the filter worked. But when the journalists asked for a “video game plumber,” the AI allowed the request and generated several very recognizable versions of the mustachioed character in a red jumpsuit. Following this incident, Autodesk stated that it had strengthened its filtering, but admitted that their models are trained on a combination of proprietary and third-party data, and the risk of generating copyrighted objects cannot be completely eliminated. The company cites its “Trusted AI Framework,” which is intended to reduce risks but does not guarantee their absence, and the ultimate responsibility for using generated models lies with the user.

For the film industry, Wonder 3D means the ability to quickly create rough models of characters and background objects — Autodesk positions the tool as a solution for rapid prototyping and pre-visualization of scenes. Instead of days of modeling, you get minutes of generation and hours of refinement. For gadgets and 3D printing, the OBJ export opens the door to rapid prototyping of parts. This is not yet a revolution — the topology of the generated models is not ideal for animation, but the company has promised to add rigging (a skeleton) in the next version. Wonder 3D is the first practical tool that allows you to “draw with words” a 3D model in a minute. And although it has already managed to “draw” Mario in defiance of its own filters, the very existence of this technology changes the rules of the game for everyone working with 3D graphics, film, and design.