ByteDance · video
OmniHuman 2
ByteDance's newest talking-photo model: natural expressions and body motion for 4 seconds to 5 minutes of audio.
Talking photo
- References
- Up to 2 image or audio files
About OmniHuman 2
OmniHuman 2 is ByteDance's newest talking-photo model, priced at $0.27 per second of audio. You give it a picture of a person and an audio file, and it returns a video of them speaking, with expressions and movement that follow the voice.
It accepts audio from a few seconds to several minutes, much longer than the earlier version, and has no resolution setting. A typical generation takes about 5 min.
Good for
- Long talking-photo videos, several minutes from a single upload
- Natural expressions and body motion driven by the voice
Getting good results
- The photo can be JPEG, PNG, WebP, BMP, TIFF, GIF, HEIC or HEIF; the audio WAV or MP3.
- The provider needs a few seconds of audio at least; very short clips are refused.
OmniHuman 2 questions
What does a minute of video cost?
$15.75, against $24.00 on OmniHuman 1.5.