Private prototype · Face-only rig · Version 6

🦄 Sunshine's Mouth Lab 6

Now Sunshine’s face and chin move independently from her mane. A hand-traced, feathered face layer limits the jaw deformation to skin pixels, so the surrounding curls and hair stay still. Your working Little Talker 1.9 remains untouched.

Ready. Your proven mouth settings are loaded.

Step 1 · Pick a recording

Choose an MP3, WAV, M4A or MP4 with Sunshine speaking

Your recording stays on this device. The page does not upload it anywhere.

Tip: use a clear speaking clip rather than a clip with music, as music will also move Sunshine's mouth.

Step 2 · Refine Sunshine's expression

Your settings are loaded: X 53.2 · Y 43.8 · size 39 · sensitivity 8 · smoothness 6 · smile 6 · muzzle 6. New: isolate the face from the hair so only her lower face changes shape. The surrounding mane stays fixed. Changes are saved only on this device.

Mouth appearance

Smile curve (upturned corners)

Nose & muzzle movement

Real chin / face-outline movement

Cheek-to-chin connection

Show jaw curve and vector control points Show face boundary (green) — check hair stays outside

The green line marks the area allowed to bend. If part of Sunshine’s mane still moves, send me a screenshot with this line showing and I can tighten the outline.

Play a voice clip or the silent preview. Try chin movement 4, then 7 to see the actual face contour change. The vector points show the jaw outline being moved.

Mouth position — left / right

Mouth position — up / down

Mouth size

Reaction strength

Smoothness

This is the fifth visual experiment: Unlike version 4, this version deforms the portrait pixels around the original jaw line with a cheek-to-chin mesh and overlays a matching curved vector outline. It is a real image-warp prototype, though we still need rigged layered artwork for perfect 3D-quality movement. This is volume-responsive, not word-accurate lip-sync. Voice files stay on the device. The facial blend uses Safari background blur and may look slightly different on other browsers.

What comes after this?

If this jaw-contour motion is on the right track, we can refine the deformation and eventually produce properly layered character artwork for true phoneme-based lip-sync. Exact word-accurate lip-sync still needs phoneme/viseme timing.