Chinese ByTedance revolutionizes video production with artificial intelligence

Chinese ByTedance revolutionizes video production with artificial intelligence

by
Chinese ByTedance revolutionizes video production with artificial intelligence





Chinese ByTedance revolutionizes video production with artificial intelligence



See in full size


Teknoloji devi Bytedanceintroduced the new artificial intelligence model that breaks ground in the field of human animation: Omnihuman-1. This innovative system, only A single image and movement signal (audio, video or both) using extremely realistic human videos can create.

The recent development of artificial intelligence technologies is now unimaginable. In addition to new scientific discoveries, the world goes towards a different path, as productive models now produce extremely realistic videos. If the use potential of these vehicles is open, it is unlimited. At least ByTedance’s Omnihuman-1 model points to this.

Artificial intelligence that brings realism to the top

Omnihuman, to increase the naturalness of movement and visual details Multi -mod movement conditioning (Multimodal Motion Conditioning) adopts the strategy. This approach offers a system that can produce real people’s movements by overcoming the difficulties faced due to the lack of high quality data.

Omnihuman is extremely extremely based on weak signal inputs, especially sound. realistic human videos It performs significantly better than the existing methods by producing. Portrait, half -body or full body images, supporting the image inputs of each attempt to support more realistic and higher quality results in various scenarios.

https://www.youtube.com/watch?v=XF5vOR7Bpzs

Omnihuman, especially Singing and Speaking such scenarios. While it can produce gestures and body movements suitable for different music genres, it offers close to real results with high -act songs and complex mimics. In the upper and below video, you can watch Nvidia CEO Jensen Huang and Taylor Swift singing in Chinese and Japanese language.

https://www.youtube.com/watch?v=GGTcVOb2S9k

Omnihuman is not only limited to human images. Animation characterseven for artificial objects and even animals, it can create real movement animations. Especially complex poses and detailed hand movements Successful results are obtained on issues where existing systems are forced. The model can also copy video movements and mimic a certain body area or general body movements. This feature can open new doors in virtual avatars and digital human projects. You can watch the video I put together with various examples below.















Download DH iOS application