DeepMotion

Create character motion with DeepMotion

DeepMotion provides AI tools for creating 3D human animation. Animate 3D extracts motion from video, while SayMotion generates motion from text prompts. Both support workflows that move beyond a rendered clip into animation data for use with a character and a separate 3D application.

It can suit an independent game developer blocking a character action, an animator adapting a recorded performance or an educator demonstrating motion capture. The distinction between motion and a complete scene matters: generated movement still needs an appropriate character, camera, environment and integration into the final project.

Video capture, pose correction and text prompts

Animate 3D accepts a source video and applies the extracted movement to a chosen character. Its current product page documents multi-person tracking, custom FBX, GLB or VRM characters, and options including face and hand tracking, foot locking, motion smoothing and physics simulation. Availability and usage depend on the chosen settings and account.

The Rotoscope Pose Editor lets you refine the character's pose over the original video. This offers a correction step when automatic tracking misses a joint position. It should be part of evaluating a demanding clip rather than assuming the first processed result needs no cleanup.

SayMotion starts with a text description and selected character. Its inpainting workflow can add, extend or blend movement using prompts. This is useful for exploring an action without a filmed performance, but a requested movement is a creative direction rather than a guarantee of precise choreography.

Example: block an action for a game character

Suppose a small game team needs a character to step forward, wave and return to an idle stance. Record a short permitted performance with the full body visible and a stationary camera. Keep the beginning and end clear so the motion can be inspected without an abrupt cut.

Upload the video to Animate 3D, choose a default character initially and inspect the feet, shoulders and hand position in the preview. Use pose correction where necessary. Then try the movement on the intended custom character and examine whether its proportions change the result, such as a hand passing through a larger torso.

For a variation that has no reference performance, describe the action in SayMotion and review its timing and style. Keep alternate versions separate instead of assuming they will blend cleanly. Export animation data to the target 3D application, check scale and skeleton mapping, and refine transitions into the game's idle animation. A preview render can help teammates discuss the action, but it is not the same deliverable as editable skeletal data.

Capture and export limitations

DeepMotion's capture guidance recommends a fixed camera, clear lighting, visible joints and an unobstructed full body. Motion blur, camera movement, cropping and occlusion can weaken tracking. Choose a controlled sample before attempting complicated overlapping performers.

The export documentation distinguishes skeletal formats such as FBX and BVH from rendered MP4 video, with GLB/GLTF and character-import details also documented. Check the particular tool, custom-character requirements and destination rig. Export compatibility does not guarantee that retargeting, contacts or cloth behavior will be finished automatically.

Free evaluation, credits and commercial licensing

Animate 3D pricing provides a free account for personal, noncommercial use and paid subscriptions. Animation credits measure processing, with additional settings affecting consumption. Its FAQ states that commercial licensing depends on the plan, including when credits were earned through labeling or correction tasks.

SayMotion has separate plan information; confirm its current allowance and intended-use terms rather than transferring Animate 3D entitlements across products. API and real-time SDK arrangements are also separate from a browser animation account.