Cartwheel: The Essentials

Last updated: July 21, 2026

The Cartwheel family brings 3D character animation to Scenario, turning images, meshes, text prompts, and reference video into animation-ready, rigged characters you can export straight into your engine or DCC tool. It covers four models. Three build and animate rigged 3D characters: Cartwheel Character Rigging, Cartwheel Text to Motion, and Cartwheel Video to Motion. The fourth model, Cartwheel Mascot, works differently: it swaps the characters in an existing video for one you define, a video-to-video edit covered at the end of this article.

What It Does

Auto-rigs characters, generating a clean skeleton from a character image or an existing 3D mesh, with no manual bone placement.
Creates motion from a text prompt, describing an action and applying it to your own character.
Captures motion from video using markerless mocap, then retargets it onto your character.
Exports production-ready files as GLB or FBX for Blender, Maya, Unreal, and Roblox, plus MA and MB.
Fine-tunes the output with controls for frame rate, orientation axes, looping, move-in-place, keyframe cleaning, facial expression, and hand pose.

Cartwheel Mascot

Cartwheel Mascot extends the family beyond 3D animation into finished video. Instead of rigging and animating a character, it takes an existing clip and swaps the people or characters in it for a mascot or custom character you define, keeping the original motion, timing, and camera. This is a video-to-video edit, distinct from the rigging and motion models above.

What It Adds

Character replacement in real footage: upload a source video (MP4, WEBM, or MOV) and the performers in it are re-skinned as your character while their movement is preserved.
Reference-image driven: supply one or more Character Reference Images that define how the replacement looks. Order matters, the first image maps to the first character in the video, the second to the next, and so on.
Prompt guidance: a Prompt describes how the new character should look and behave in the scene, refining the swap.
Optional environment: an Environment Image can set the background or setting the character appears in.

Key Parameters

Provide an Input Video (required) as the clip to transform, and one or more Character Reference Images (required, ordered per character) showing what the replacement should look like. Write a Prompt (required, up to 4096 characters) describing the new character and how it should appear. Optionally add an Environment Image to control the background or setting.

Examples

A performer in the source clip replaced by a pink rabbit mascot in a bow tie · Open on Scenario

A character swapped for a cartoon figure in a uniform, original motion preserved · Open on Scenario

A performer restyled as a toy fox mascot in a jacket and boots · Open on Scenario

Where It Fits

Reach for Cartwheel Mascot when you already have a video and want to re-cast who or what is in it, for example turning live footage or an existing animation into branded mascot content. Use the Character Rigging, Text to Motion, and Video to Motion models instead when you need to build and animate a rigged 3D character from scratch rather than edit an existing clip.

Which Model Should I Use?

Model	ID	Best for
Cartwheel Character Rigging	model_cartwheel-character-rigging	Turning a character image or raw mesh into a rigged, animation-ready model.
Cartwheel Text to Motion	model_cartwheel-text-to-motion	Generating animation from a written description, no reference footage needed.
Cartwheel Video to Motion	model_cartwheel-video-to-motion	Copying real movement from a video onto your character, with optional facial capture.
Cartwheel Mascot	model_cartwheel-mascot	Replacing the characters in an existing video with a mascot or custom character, keeping the original motion.

Parameters

Cartwheel Character Rigging

Provide either a Reference Image of your character or a 3D Model (FBX, GLB, or OBJ). If both are given, the 3D model takes precedence. The output is a rigged mesh with a clean skeleton, ready to receive motion from the other two models.

Cartwheel Text to Motion

The key input is the Prompt describing the movement, applied to a supplied Character 3D Model or Reference Character Image. Tune Duration (3 to 8 seconds) and CFG Scale for prompt adherence. Export controls include Output Format, Frame Rate, Forward and Up axes, Move In Place, Loop, Keyframe Cleaning, Face Expression, and Hand Pose.

Cartwheel Video to Motion

Supply a Reference Video (up to 30 seconds; MP4, MOV, AVI, MKV, or WEBM) plus a Character 3D Model or Reference Character Image. Set People (up to 4) to track multiple performers, choose the Motion Capture Pipeline (Comic 4 or Comic 3), and enable Facial Capture on Comic 4. The same export controls as Text to Motion apply.

Examples

Character Rigging: a reference character auto-rigged into an animation-ready model · Open on Scenario

Character Rigging: a detailed mesh prepared with a clean skeleton · Open on Scenario

Text to Motion: a prompt-driven animation applied to a stylized character · Open on Scenario

Text to Motion: an energetic action generated from a text description · Open on Scenario

Video to Motion: body movement captured from footage and retargeted · Open on Scenario

Video to Motion: markerless mocap driving a different character · Open on Scenario

Tips for Better Results

Rig first: run Character Rigging (or let Text to Motion and Video to Motion auto-rig) before expecting clean animation, since motion needs a proper skeleton.
Use a clear, full-body reference image or mesh in a neutral pose so the skeleton is placed accurately.
For Text to Motion, describe one continuous action clearly, for example "walks forward, then jumps and waves", and raise CFG Scale when you want stricter adherence.
Keep Text to Motion durations within the 3 to 8 second range and consider Loop for cycles like walking or idling.
For Video to Motion, use steady footage with the full body visible, and set People to match how many performers you actually need to track.
Match Forward and Up axes to your target software (Blender, Maya, Unreal, or Roblox) if the motion imports facing the wrong way.
Enable Move In Place for game engines where locomotion is handled by code, and use Keyframe Cleaning to lighten files for hand-editing.

Known Limitations

Text to Motion clips are limited to 3 to 8 seconds per generation.
Video to Motion accepts up to 30 seconds of footage and tracks a maximum of four people.
Facial capture is only available with the Comic 4 pipeline in Video to Motion.
IK feet and IK hands options apply only to Maya exports (FBX Maya, MA, MB).
Results depend on input quality: unclear reference images, cluttered meshes, or shaky, occluded video can reduce rig and motion accuracy.

Use Cases

Games: quickly rig concept characters and build motion libraries for prototypes and production.
Film and animation: block out shots and generate character performance without a mocap stage.
Marketing: animate mascots and branded characters for ads and social content.
Product and virtual production: turn reference footage into reusable, retargetable animation for demos and previz.