VRM mouth not moving? Your model may have no mouth shapes

The avatar loads, the head follows you, the body moves - and the mouth stays shut no matter how loudly you talk. Sometimes that is a setting in your VTuber app. Far more often, especially with AI-generated or converted models, the model simply has nothing to move its mouth with.

If your VRM's mouth does not move when you talk, check two things in order. First the app: lip sync switched on, and the right microphone selected. Then the model: a VRM lip syncs through five mouth expressions (aa, ih, ou, ee, oh), and if they are missing or empty, nothing will ever move. Models from AI generators and auto-riggers like Meshy, Tripo and Mixamo have no mouth shapes, and converters cannot add what is not there - so their VRMs load fine and never speak. The fix is to build the shapes: upload the model to Riggle and download a VRM 1.0 that talks and blinks.

Fix my avatar's mouth

Rule out the app first

  • Is lip sync on? Most VTuber apps have a separate lip sync or audio setting, sometimes off by default.
  • Is it hearing you? Check the microphone input in the app, not just in your system settings, and watch its level meter while you speak.
  • Is it the right kind of lip sync? Voice-based lip sync drives the five vowel expressions. Camera-based face tracking with an iPhone usually wants the 52 ARKit "perfect sync" shapes instead - a different set, and a different problem: perfect sync not working.
  • Does the app read your VRM version? VRM 1.0 and VRM 0.x are different formats, and some apps read only one of them.

If another VRM - a sample model, a VRoid export - lip syncs fine in the same app with the same settings, the app is not the problem. The model is.

Check whether the model has mouth shapes at all

A VRM moves its mouth through expressions named aa, ih, ou, ee and oh - one per vowel. Each expression points at a shape on the mesh (also called a blendshape, morph target or shape key) that holds the mouth in that position.

Open the VRM in Blender with the VRM add-on and look at its expressions, or load it in any VRM viewer that lists them. You will find one of three things:

  • The expressions are there and point at shapes. The model is fine - go back to the app settings.
  • The expressions are there but empty - they exist, but nothing is bound to them. This is what most converters produce from a model without face shapes.
  • The mesh has no face shapes at all. Typical for Meshy, Tripo, Mixamo and marketplace models.

In the last two cases no setting will help. The shapes have to be made.

Why converted models end up like this

A GLB-to-VRM converter does two jobs: it maps the skeleton to VRM's humanoid bones, and it carries over the face shapes the model already has. AI generators and auto-riggers rig the body - arms, legs, spine, head - and stop at the neck. There is nothing to carry over, so the converter writes a valid VRM with empty mouth expressions, and it loads everywhere without a single error.

The fix: build the mouth shapes

Building the shapes by hand means sculpting each vowel in Blender, then wiring them into the VRM in Unity - an afternoon at least, per model. With Riggle it takes about a minute:

  1. Upload the .vrm you already have, or the original .glb.
  2. Watch a video of your avatar saying each vowel, blinking and smiling.
  3. Download it.

Riggle shows you your model's face and asks you to point out the mouth and eyes - a few clicks. You get back a VRM 1.0 with the five mouth shapes, blinking (both eyes and each on its own), happy, angry and surprised, all shown in the preview before you pay. The skeleton you already had is kept. More on the conversion.

Make your avatar talk

Upload the VRM that will not talk - or the GLB it came from - and watch it speak before you download. Your first model each month is free.

Fix my avatar's mouth

FAQ

Why does my VRM blink but not talk?

Blinking and talking use different expressions, so a model can have a blink shape and no mouth shapes. Check that aa, ih, ou, ee and oh exist and are bound to shapes on the mesh.

Can I add mouth shapes without Blender or Unity?

Yes. Upload the model to Riggle in the browser and download a VRM with the shapes built in. Nothing to install.

Will Riggle keep my existing rig and textures?

Yes. Riggle keeps the skeleton, the mesh and the materials, and adds the face expressions. The output is VRM 1.0.

The mouth moves, but it looks wrong. What now?

If the shapes exist but stretch the whole lower face, upload the model to Riggle: it rebuilds the mouth shapes, you watch every one in the preview video before you pay, and re-running is free until it looks right.

How much does it cost?

Uploading and previews are free. Downloading a finished VRM costs 10 tokens once per model, and the free plan includes 10 tokens a month.

Guides