Expression guide
Give your avatar expressions in Virmii.
Virmii now supports AI expression generation and binding. Generate expression artwork from your model’s face and use it with camera tracking, including intermediate blink and mouth poses. No manual Shape Key binding is required for this workflow.
- 1.Front-facing model reference
- 2.13 expression images
- 3.Camera-driven playback in Virmii
The 3D model still moves with your head and body. Expression artwork follows the face; smaller mouth poses and half-blinks provide intermediate steps.
Open your completed model
Open a completed Virmii-generated model and select Generate expressions. Without a subscription, each set costs 20 credits. Creator and Pro use included monthly generations first, then 20 credits per set. Subscribers can see their remaining included generations at the top of the model’s AI expressions panel. This allowance is shared across all their models.
Virmii captures a neutral, front-facing render of the model and sends that reference to Gemini. Your webcam is not needed for generation, and no webcam image is sent in this step.
Wait for the expression artwork
Virmii generates 13 images: closed and half-closed eyes, a smile, five vowel mouth shapes, and five smaller mouth shapes. The original face provides the neutral pose.
Background generation may queue for up to 24 hours. You can leave the page and return later. The completion badge appears after all 13 images are ready.
Retry only what is missing
If generation fails, the initial charge is kept. Each model version allows up to 2 free retries for missing or failed images. Further retries are blocked after both are used. Completed images stay unchanged, and retries use no additional credits, included generations, or model retry attempts.
Once an expression image succeeds, it is fixed for that model version and cannot be regenerated. Retrying does not replace successful images or create a new full set.
Test expressions and go live
After completion, enable camera tracking in the model preview or Live Studio. Test each eye separately, gentle and wide mouth openings, and head movement.
Adjust blink and mouth sensitivity to match your movements. Live Studio settings are saved and shared with the OBS stage; preview adjustments are temporary. Use your OBS live URL as a Browser Source.
Where these expressions work
Expressions play inside Virmii
The generated artwork is stored separately from the model and used in Virmii previews, Live Studio, and the OBS stage. Downloading the VRM does not embed these images as facial expressions or add Shape Keys. Other VRM apps will use the file’s original expression bindings.
Uploaded VRMs use their existing expressions
AI expression generation and binding is available for models created in Virmii. Uploaded VRMs use the blink and mouth expressions already bound in their files; uploading does not add missing expressions.
Check the result on your character
Appearance depends on the generated artwork and the model’s face geometry. Raised lips, eyelids, hair over the face, and large head turns can expose alignment differences. Intermediate images are discrete poses, not continuous mesh deformation.
Common questions
Can I use a model generated from text?
Yes. The reference is captured from the completed 3D model, so an original character image is not required. The face still needs to be recognizable from the front.
A wink does not trigger, or the mouth opens too much.
Keep your face visible to the camera. Increase blink sensitivity for subtle winks, or lower mouth sensitivity for smaller openings. In Live Studio, adjust the open-eye baseline if your eyes appear closed while relaxed.
Do I have to keep the page open?
No. Generation continues in the background. Return to the same model to see progress, retry a failed task, or start tracking once the completion badge appears.