Wednesday, March 20th, 2024 Posted by Jim Thacker

DeepMotion launches AI text-to-animation service SayMotion

AI motion-capture specialist DeepMotion has launched SayMotion, a new browser-based generative AI service for creating character animation from text prompts.

The platform, which can currently be used for free in open beta, generates animation in FBX, GLB and BVH format, for use in DCC software and game engines.

The service aims to provide an intuitive way for users to generate base motions suitable for use in games, AR, VR, animation or motion graphics projects.

Generate detailed animations of human characters by chaining up text promts
Founded in 2014 by games industry veteran Kevin He, DeepMotion is one of pioneers in the field of AI motion capture, and paved the way for other recent browser-based mocap tools.

With SayMotion, it has moved into the world of generative AI, letting users create animation clips by entering text descriptions of the movement they want to generate.

Prompts follow a three-part structure: subject, action and detail.

Subject describes the nature of the character being animated (for example, ‘woman’, ‘child’ or ‘swimmer’), and action desribes… well, the action being performed.

Detail modifies the animation, by specifying the direction or orientation in which it should take place, and the style or emotion with which the character should carry it out.

When chained together, they can generate quite complex sequences of actions.

One example from the online documentation uses the prompt, ‘A person walks clockwise in an angry manner, and then reaches out with their left arm to grab the left arm of a chair, lowering themselves heavily into the seat.

Inpainting lets users extend or refine existing animations
The initial animation can also be refined using SayMotion’s ‘inpainting’ system, which lets users extend, modify or blend animations with additional text prompts.

The example in the starter guide shows the motion of bending down to pick up an object being layered onto an animation of a character walking in a circle.

The system also supports some of the advanced options from Animate 3D, DeepMotion’s AI mocap service, including motion smoothing and foot locking.

Supports custom 3D characters, and exports in standard animation formats
As well as DeepMotion’s readymade characters, or those created with its built-in avatar-generation tools, users can upload their own custom 3D characters to animate.

Those from stock model sites like Mixamo, Sketchfab and TurboSquid should work, so long as they have standard humanoid bodily proportions.

The animated character can then be exported in FBX, GLB or BVH formats, making it possible to use the animation in DCC applications like Blender or Maya and game engines like Unity and Unreal Engine.

It is also possible to export video of the animation in MP4 format.

Current limitations
DeepMotion is currently limited to generating full-body motion of human characters, although DeepMotion is considering adding support for facial and hand animation.

The company also plans to add support for video prompts, generating a new unique animation based on SayMotion’s interpretation of a source video.

Price, system requirements and release date
SayMotion is currently in open beta. It should work with any standard desktop web browser. Support for mobile browsers is planned.

During the beta period, all users get 2,000 credits per month. Generating an animation or MP4, downloading an animation, or using inpainting all consume a credit.

Animation created with beta accounts is licensed for use in commercial projects.

DeepMotion hasn’t announced a release date or final pricing for SayMotion, but you can find details of its Animate3D subscription plans in this story.

Read an overview of how SayMotion works on DeepMotion’s blog

Visit the SayMotion product webpage

Have your say on this story by following CG Channel on Facebook, Instagram and X (formerly Twitter). As well as being able to comment on stories, followers of our social media accounts can see videos we don’t post on the site itself, including making-ofs for the latest VFX movies, animations, games cinematics and motion graphics projects.