Liverty five-types - raw lip-sync take, v1
The talking layer only. No captions, no image overlays, no music, no end card. 114.2 s, 464x832, 25 fps.
This is NOT the finished ad. Still to come: house TikTok captions, the image overlays (which on this
chassis are the only thing that changes the frame), and the adversarial review. It is here because you asked to see it now.
Unmute it to judge the sync.
What it is
| script | s03 five-types, donor MRP9R55DEL9FI (acne-scar diagnostic ad), 353 words, registered swiped body at ai-kill 0 hits |
|---|
| voice | w30-kristina, one of the 13 belt voices approved 2026-11-03, 185 wpm |
|---|
| presenter | a real frame from a real clip in the brick library, face changed by gpt-image-2.5 edit so she is nobody real |
|---|
| lip-sync | WaveSpeed InfiniteTalk 480p, 13 min render |
|---|
| cost | $3.48 total: $0.07 voice, $0.03 face, $3.38 lip-sync |
|---|
Verified by me: duration matches the voiceover exactly, the same woman holds across six sample
points with no drift, and real speech is present at -35 dB mean and -10.9 dB peak.
Three things I want your eye on.
1. Does the sync read, or does it look robotic? Stills cannot tell me this and it is the main reason this is in front of you.
2. She has visible facial eczema. That fits a sufferer explaining the five types, and it sits oddly under a close that sells the fix.
3. Her t-shirt carries a printed graphic. If it reads as another brand it needs covering or a different base frame.