🎨 OmniHuman 2.0: Up to 5 Minutes in a Single Take
BytePlus (ByteDance's cloud) has released OmniHuman 2.0 — a model that creates animation with lip-sync, facial expressions, and gestures from a single photo and audio track, synchronized with speech or singing. A single take has grown from 60 seconds to 5 minutes, with the addition of native 720p, gesture and camera control, and multi-character scenes. The price has dropped from $0.12 to $0.07/second.
🌍 A long take without splicing and a $0.07/second rate make avatars profitable for advertising, e-commerce, music videos, and localization, where videos were previously assembled from short pieces with visible seams.
👤 The launch is available via the BytePlus Console and API: 30 seconds — $2.10, 5 minutes ≈ $21 per attempt, and failed takes multiply costs. The 5-minute distance in the announcement is not shown; videos with a real face require the person's consent.
Source 1: https://www.byteplus.com/en/product/OmniHuman Source 2: https://docs.byteplus.com/en/docs/byteplus-vision/omnihuman-2-0-video-generation