Metlivi Blog

How to Edit Lip Sync and Body Animation on Separate Bone Channels

To edit lip sync without disturbing body animation, give each performance its own action containing only the properties it should control, then play those actions together on separate animation tracks. Keep jaw and lip controls in the mouth action; keep root movement, limbs, torso, neck, and head turns in the body action.

This walkthrough uses Blender 4.5 LTS and a character that already has working mouth and body controls. The concrete task is to adjust mouth poses and dialogue timing while preserving an existing body performance. Blender’s Nonlinear Animation editor, or NLA editor, provides the tracks for combining those actions. Blender Manual: Tracks

October 29, 20268 min readBy Metlivi Editorial Team
Section 1

What actually separates mouth animation from body animation?

In Blender, an action stores animation data. Its F-curves describe how individual properties change over time, such as a bone’s rotation or a control’s custom property. The channel groups displayed under bone names help organize those curves, but their names do not establish exclusive control over a character part. Blender Manual: Actions

The practical rule is to assign each animated property one owner. Two actions named Body and LipSync can still interfere if both contain curves for the same jaw control. Conversely, actions on one armature can work together cleanly when their animated properties are separate. This workflow follows from separating the curves themselves.

Action slots serve another purpose: they organize distinct animation data for data-blocks within an action. For this single-armature workflow, use separate body and mouth actions, each with an appropriate slot assigned to the armature object. Creating slots named “mouth” and “body” does not automatically filter bones into those categories. Blender Manual: Action Slots

Section 2

Which controls should belong to each action?

Before editing, make a small ownership map. The following is a suggested arrangement, not a required Blender naming convention:

Control or property Suggested owner Editing rule
Root movement, torso, limbs Body action Preserve during dialogue revisions
Neck and head turns Body action Keep the character’s overall performance together
Jaw opening and lip controls Lip-sync action Adjust against the dialogue
Blink and brow controls Optional expression action Separate only if independent editing is useful
A custom property driving mouth shapes Lip-sync action Key the property that actually produces the shape

Test the rig before recording keys. Move one mouth control, observe what changes, then restore it. Identify whether the visible shape comes from the control’s transform, a custom property, or a driven relationship. Blender’s final animation can include drivers and constraints as well as keyframed transforms, so an apparently empty deformation-bone channel does not establish that the mouth has no animation. Blender Manual: Editing Strips—Bake Action

Keep the existing hierarchy intact. Child bones normally inherit their parents’ position, rotation, and scale. Consequently, mouth controls can have independent local animation while still following a head turn. There is usually no reason to disconnect the jaw or change bone parenting just to separate editing. Blender Manual: Relations

Section 3

How do you create a mouth action beside an existing body action?

Start by saving a working copy of the project. Select the armature, open the Dope Sheet’s Action Editor mode, and inspect the existing body action. Expand its channels and check that it contains the properties assigned to the body in your ownership map.

Use Push Down Action to move that action into an NLA strip. This leaves the armature without an active action. Blender distinguishes this from Stash Action, which places the action on a muted track for later use. Use Push Down when the body should keep playing while you work. Blender Manual: Action Editor

Create a new action in the now-empty Action Editor and name it clearly, such as LipSync_Line01. Enter Pose Mode, select a mouth control, and insert an initial key on its intended property. Confirm that the new action and its slot are assigned to the armature before continuing.

Key only the properties on your mouth list. For repeated work, create a custom keying set containing those properties. Keying sets insert keys for a defined collection of properties, so check the active set before each pass; a broad character set may include more controls than this task needs. Blender Manual: Keying Sets

After establishing the mouth animation, push it down too. Put its strip on a separate NLA track above the body track, with both strips covering the intended performance interval. You now have separate clips to select, inspect, and revise.

Section 4

How do you split an action that already contains both performances?

If the original action already includes mouth and body keys, preserve it before removing any channels. A practical approach is to push it into the NLA editor, duplicate its strip with Strip → Duplicate, and place the copy on another track. This command duplicates the referenced action; Linked Duplicate shares it. Make Single User can separate an existing shared action before editing. Blender Manual: Editing Strips

Rename the two actions Body and LipSync. Edit the body copy and remove only the mouth-owned channels. Then edit the lip-sync copy and remove the body-owned channels. Retain an untouched project copy so that an overlooked control can be recovered.

Remove whole unwanted channels using Channel → Delete Channels. With keyboard shortcuts, the pointer must be over the channel region; over the keyframe area, deletion affects selected keys instead. Inspect property paths where a control has several responsibilities, rather than deleting a whole bone group without checking its contents. Blender Manual: Editing Channels

Test each result in isolation. The body action should reproduce the body performance without the recorded mouth motion. The mouth action should contain the dialogue shapes without root movement or gestures. Then enable both to check the combined result.

Section 5

Which blending settings keep the actions independent?

For two actions with separate property ownership, start with Replace blending and full influence. Replace overwrites lower-track values for properties animated by the upper strip; it does not create a bone filter. If enabling lip sync changes a body pose, inspect the mouth action for unwanted body curves before changing blend modes. Blender Manual: NLA Sidebar

Set the mouth strip’s extrapolation to Nothing when it should contribute only within its boundaries. Hold and Hold Forward can extend endpoint values beyond the clip. Also key an intentional resting mouth pose near the start and end if you need a controlled transition; extrapolation alone does not design that transition.

Treat these as different controls: ownership determines which properties a clip animates, blending determines how overlapping values combine, and extrapolation determines what happens outside the strip. Diagnose them in that order.

Section 6

How do you adjust lip-sync timing without shifting gestures?

Select the mouth strip and enter Tweak Mode with Tab in the NLA editor. Blender’s full-stack tweak mode keeps the other tracks enabled while exposing the selected action’s keys. That lets you refine mouth motion while seeing the body performance. Exit Tweak Mode before moving the strip itself. Blender Manual: Start Tweaking Strips Actions

Choose the smallest timing edit that solves the problem. Move individual mouth keys when one shape arrives early or late. Move the entire mouth strip when the complete line has a consistent offset. Scale the strip only when its overall duration needs to change. Blender distinguishes moving a strip, cropping its action range, and changing its playback scale. Blender Manual: NLA Sidebar

For an illustrative calculation, suppose a project runs at 24 frames per second and every mouth shape starts 0.25 seconds late. The offset is 24 × 0.25 = 6 frames. Move the mouth strip six frames earlier, leaving the body strip and audio reference in place. If the beginning aligns but the ending drifts, a single offset will not fix the whole line; inspect duration and intermediate poses instead.

After each revision, replay the complete line and inspect its first and last mouth poses. A local correction can improve one syllable while making the transition beside it abrupt. Keep the body track visible for this final pass.

Section 7

How can you diagnose interference quickly?

Use this sequence to locate the cause before rebuilding the animation:

  1. Inspect channel ownership. Look for root, limb, head, or unrelated custom-property curves inside the mouth action.
  2. Solo each track. Blender’s Solo control isolates a track, including muting the active Action Track. Confirm that each clip contains the intended performance. Blender Manual: Tracks
  3. Check shared actions. If changing one strip changes another, inspect whether they reference the same action; make an independent copy when necessary.
  4. Inspect strip boundaries. For a mouth pose that persists after dialogue, check extrapolation and endpoint keys.
  5. Inspect drivers and constraints. If curves are separate but the rig still couples movements, trace the controls that produce the final pose.

Do not use channel locks as a substitute for separation. A protected channel cannot be edited, while a muted channel stops affecting animation. Neither operation removes that channel from the action. For a reusable mouth-only clip, clean ownership is easier to inspect than a collection of temporarily disabled curves. Blender Manual: Editing Channels

Section 8

How do you preserve the combined result when exporting?

If the task ends with a Blender render, retain the separate actions for later revisions. If you need a glTF character file, choose the export mode deliberately: the exporter’s NLA Tracks mode exports each track as an independent animation, whereas Scene mode exports the evaluated scene animation. Two visible tracks therefore do not automatically mean one combined exported clip. Khronos glTF-Blender-IO documentation: Animations

For one combined performance, use an export configuration intended to capture the evaluated scene, then inspect the exported file in its destination. Check dialogue timing, head movement, mouth endpoints, and the selected frame range. The exporter documentation also specifies that sampling bone-driven shape-key animation requires the mesh to be a direct child of the armature. Check that condition if your mouth rig uses that arrangement. Khronos glTF-Blender-IO documentation: Export

The finished setup should let you revise a mouth pose, shift a dialogue clip, and adjust a gesture independently. Keep the editable actions in the source project, verify their property ownership, and review their combined playback after every timing change.

Section 9

Sources

Related reading

Keep exploring this topic