Tim Simmons reviews ByteDance's OmniHuman avatar examples and a world-model demonstration involving camera changes and interactive scenes. The account is explicitly limited: OmniHuman was region-locked for the presenter, and the world-model footage does not establish the final interface or public capabilities. The discussion treats apparent progress and remaining lip-sync problems as observations, not a completed independent evaluation.
Tim Simmons then uses Krea 3 editing examples to explore texture cleanup, localized changes and continuity between animation frames. The emphasis is on targeted editing rather than assuming one model is best at every generation or compositing task. A separate Sora segment describes reported talks about professional access to related technology, not a confirmed consumer relaunch. The closing procedural-city example is an additional creative workflow, with export possibilities identified as untested.
Watch on YouTube
