Lip sync calibration

Accepted live preset: Balanced + timing compensation (v1). Keep saved clips and review notes here for future calibration.

Checking access…

Voice sample

Creating a clip uses the normal AI and voice budget once. Replays and preset switches make no AI requests. Clips stay in this browser until cleared.

LAB-001 · timing compensation Reference · Previous Balanced Accepted · Live v1

Same movement strength and smoothing. Candidate anticipates eligible mouth targets by 35 ms to compensate for blending lag. It does not change the voice or fix pronunciation guesses.

No clip yet.

Inspect a specific moment

Playback speed (slower playback lowers pitch)

Clip position

Letters and target shapes are estimates, not recognized phonemes. Paused inspection reconstructs the smoothed pose; it does not snap to the target.

Import saved benchmark

Review resultYour preference / notes

Notes save in this browser. Export the clip and review to preserve the exact sample and settings; the accepted v1 is now used by the live interview; future experiments stay here until approved.

Generation timing

Available after creating a clip. Preparation time here includes saving the whole clip; it does not measure live interview startup.