An audible soundtrack does not prove that a puppet has recognized mouth shapes available to show. In Character Animator, inspect the Mouth group's artwork and tags before regenerating a correct voice. Adobe documents named viseme states and the Lip Sync behavior that selects them; a single painted mouth supplies no alternatives even when audio reaches the scene.
Separate voice, analysis and visible artwork
Verify the spoken recording independently first. Then inspect whether a Lip Sync take or live input is actually driving the intended puppet. Finally, inspect the drawings that the resulting states can select. These checks separate a missing recording, an undriven behavior and a rig containing only one usable mouth.
Avoid increasing volume blindly. Loudness cannot create an absent closed-mouth drawing or correct a state linked to the wrong layer. Likewise, an animated take can exist while several states all reveal identical artwork, making a functioning control appear visually frozen.
Make a mouth-state inventory
Open the intended Mouth group and compare its structure with the documented preparation route. Check names or explicitly assigned tags, not just what the layers resemble in a thumbnail. Confirm that the selected state displays the intended drawing and that unrelated variants do not remain stacked over it.
Use an original planning inventory with a rest state, a clear closure and visibly distinct open shapes appropriate to the rig. This is not a prescription to invent anatomical detail from a real person or a guarantee that every sound maps perfectly. The objective is to establish that different recognized states can produce different approved pictures.
Test one short authorized phrase
For a hypothetical mascot saying an agreed greeting, preserve the recording and test a short copy. Check one place where the mouth should close, one clearly open portion and a pause. No lip-sync computation was executed for this example.
| Observation | Inspect next |
|---|---|
| No relevant take or live input | Behavior and actual input route |
| State labels change, picture does not | Drawing/tag mapping and duplicate artwork |
| Correct closure, wrong resting smile | Neutral state and competing face control |
| Mouth moves but speech looks wrong | Timing and the selected performance |
When reviewing a recorded take, prevent an unintended live camera or microphone input from obscuring what was recorded, using the current supported input controls. A smile caused by live face tracking and a phonetic mouth selected from audio are different influences. Restore the intended setup after the diagnostic.
Accept a readable performance, not changing labels
Watch the character at its final viewing size, then inspect the mouth close-up. Reject doubled outlines, unexplained jumps or states that drift away from the approved facial design. Review entry to silence so the character does not hold an accidental speaking mouth through a long pause.
Retain the mouth inventory, the audio version and the actual take used in the accepted project. Imagild can contribute an authorized character design, but it does not supply Adobe's viseme recognition or a completed rig. The handoff passes when the existing sound, selected states and visible drawings correspond through the agreed phrase; a moving meter or a successful analysis status alone is insufficient.