Avatar Stream Test

Connection

Audio Input

Recording from microphone...

Drop audio file or click to upload

Audio is sent fully, then HuBERT extracts features in one pass for consistent quality.

Parameters

Target FPS 25

JPEG Quality 65

EMA Smoothing 0

0 = no smoothing, higher = smoother lips but more delay

Film Grain 4

Cinematic grain to mask generation artifacts (0=off)

Lip Sharpen 0

Unsharp mask + CLAHE on generated face (0=off, 1-2=subtle, 3=max)

Face Blur 0.5

Gaussian blur on full frame to equalize sharpness (0=off, 1-2=subtle)

Temporal Stabilization

TMix Frames 0

Temporal frame averaging buffer (0=off, 3=recommended, 5+=aggressive)

Blend Mode paste

How the generated face is composited back onto the body frame

Output Scale 1.0

Downscale output stream (1.0=native from --target_height, 0.5=half, 0.25=quarter)

⚡ Avatar Stream