Each model received the same portrait and the same recording, on the same day, with no retries and no hand picking of takes. The face and the voice are both synthetic, so no real person is being put through eight models. The script runs 9.5 seconds. Prices are for a round 10 seconds, because two of these models charge by 10 second duration steps rather than per second of audio, and a round figure is the only basis all eight share. Each is priced at the resolution shown on its card: the models do not share a resolution ladder, and comparing a 480p price with a 1080p one would say nothing.
๐ Tap for sound
InfiniteTalk Fast
20 credits ($0.40 at the API rate)
Resolution
480p
Rate
2 credits per second of input
Length
As long as the audio
one resolution only (480p); soft image (native 352x624, upscaled to 720x1280)
Credits for one 10 second clip, cheapest first. Dollar figures are those credits at the Percify API rate of $0.02 per credit. App plans buy credits at their own rates, so treat the credit column as the comparable one.
Model
Released
10s clip
At API rate
Resolution
Length
InfiniteTalk Fast
Nov 2025
20 credits
$0.40
480p
As long as the audio
InfiniteTalk
Aug 2025
60 credits
$1.20
720p
As long as the audio
H3 Max Lip-Sync
Sep 2026
96 credits
$1.92
768P
As long as the audio
OmniHuman 2.0
Oct 2026
100 credits
$2.00
720p
Set per run, priced at 10s here
Wan 2.7 (Image to Video)
Apr 2026
100 credits
$2.00
720p
Set per run, priced at 10s here
Kling AI Avatar v2 Pro
Dec 2025
112 credits
$2.24
1080p
Set per run, priced at 10s here
Sync-3 Avatar
Apr 2026
135 credits
$2.70
1080p
Set per run, priced at 10s here
Wan 2.7 Pro (Image to Video)
Jun 2026
140 credits
$2.80
1080p
Set per run, priced at 10s here
Lip sync engines charge for every second of audio you give them, so a longer recording costs proportionally more and the figures above scale with it. Every price here is read from the same public catalog the Percify API and the model pages serve, so it moves when the product moves. See every model in the catalog or read how talking photos work.