How to Edit Motion Data by Chatting in Motion Editor

Overview

Motion Editor is QuickMagic's new motion editing feature. It imports both video motion capture (Video-to-Mocap) results and text-to-motion (Text-to-Motion) animations from the platform, and lets you refine them through natural-language conversation with an AI Agent — no manual keyframing or curve editing required.

Getting Ready

1Log in and prepare motion data

Log in to QuickMagic and prepare an existing motion asset: either a video mocap result or a text-to-motion result.

2Import into Motion Editor

Motion Editor accepts two types of motion data sources. Both lead into the same conversational editing workflow:

3Get to know the interface

The preview viewport on the left plays the character animation; the chat panel on the right is where you talk to the Agent (input at the bottom); the timeline below provides a frame ruler, playback controls, Go to / Start / End frames and FPS settings. The loop is: send instruction → Agent processes → preview and verify.

Importing motion: click the Video to Motion or Text to Motion button next to the chat input to load motion data from the platform into the current session.

Comparison view: after each edit, the viewport shows the gray original animation side by side with the purple new animation (New Anim), so you can compare the changes frame by frame.

Note: Frame numbering starts at 0: a 120-frame clip ranges from 0 to 119. Out-of-range instructions will trigger clarification questions from the Agent.

Demo 1: Blend a Body Part Smoothly

Found a stiff body-part movement? Name the limb and frame range, and the Agent will blend it smoothly, converging toward the target pose:

Blending can target a single body part (e.g. the right leg) without affecting anything else. Once finished, the Agent confirms and asks you to preview:

Processed. Motion data has been updated. Please preview the character animations. Let me know if you spot any issues for further optimization.

Check the result in the side-by-side comparison view (gray = original motion, purple = new motion). If anything looks off, keep chatting — multi-turn refinement is supported.

Demo 2: Constrain a Body Part

You can also describe the target state of a specific body part, and the Agent will adjust accordingly:

When an instruction like this has no frame range, the Agent proactively confirms the scope:

Which frame or frame range should be edited? (You can say the whole clip; saying 'whatever' means the whole clip.)

Reply 'whatever' to apply it to the whole clip — or state an exact frame range to edit only a section.

Demo 3: Extend the Clip at the End

The simplest edit: tell the Agent how many frames to extend at the end of the clip.

Both phrasings work — a shorter instruction is just as effective.

Demo 4: Generate New Motion in a Frame Range

The most powerful feature: describe the motion and specify an exact frame range — the Agent generates it in place. This is perfect for filling the extended blank frames:

The more specific the description (action name, limb order, start/end frames), the more controllable the result.

Demo 5: Smooth a Motion Transition

When the newly generated motion doesn't connect naturally with the original clip, let the Agent smooth out the transition, converging toward the target frame:

More Common Operations

Insert Frames at a Specific Position

If the requested frame is out of range, the Agent does not guess — it asks for clarification:

The clip has 120 frames (frames 0–119), so frame 180 is beyond its end. Please specify: 1) location: add the 30 frames at the end of the clip (after frame 119), or after a different frame within 0–119? 2) type: hold a still pose for 30 new frames (freeze), or insert 30 interpolated transition frames?

This shows the Agent genuinely understands the motion data structure. State the location and type (freeze frames / interpolated transition) clearly to pass in a single turn.

Delete a Frame Range

Remove an unwanted range and the clip shortens instantly — trimming motion length takes one sentence.

Replace the Action in a Frame Range

Not happy with a segment? Swap it out entirely instead of redoing it — every other frame stays untouched.

Export & Optimization

When satisfied, export in one of two ways: click Export, or simply type export in the chat — the Agent will open the export panel for you.

The export panel provides the following parameters:

• Format: target skeleton format (see the table below)
• First Frame Pose: Default / T-Pose / A-Pose, for easier downstream integration
• Displacement Data: Default / In-place
• Optimization toggles: Foot Sliding Fix / Mesh Penetration Prevention / Keyframe Reduction, combinable (details below)

Pick the skeleton format matching your downstream tool — no manual retargeting needed:

Three pre-export optimization toggles can be enabled individually or combined, applied automatically on export:

• Foot Lock: locks the planted foot, reduces foot sliding. Best for clips with locomotion or push-off.
• Auto Keyframe Reduction: slims down keyframes. Best when further manual editing in a DCC is planned.
• Anti-Penetration: reduces body-part mesh intersection. Best for clips where limbs pass close to the torso.

Prompt Tips

• Specify the frame range: from frame 35 to frame 145
• Specify the location: to the end / after frame 119
• Specify the edit type: freeze the pose / interpolated transition
• Break down the motion: first a left leg kick, then a right leg kick
• Name the body part: the right leg / the entire left arm
• Scope questions: when the Agent asks for the scope, replying 'whatever' means the whole clip
• One clear request per message: split complex edits into multiple turns

Notes

• Always preview after each edit — use the side-by-side comparison view (gray original vs. purple New Anim) to verify frame by frame before moving on
• For complex motion generation, work in segments and verify each one
• Keep a backup of the original data for comparison and rollback
• Enable export optimizations as needed — not every clip needs all three