Summary
Highlights
Bite Dance released Omnihuman, an AI tool for realistic lip-syncing and body animation from images and audio. They also released Seaweed, a new video generator model. The video tests both tools.
The Dreamina platform is used to test Omnihuman. The video demonstrates generating videos from images and both AI generated voices (text-to-speech) and uploaded audio clips. It showcases various examples with different images, including AI-generated and real photos of people like Jensen Huang and Sam Altman. Different languages were tested showcasing impressive realistic results.
Omnihuman can animate full-body photos, including details like hands and objects. Singing examples are tested with AI-generated songs, showing realistic mouth movements and body language and limited ability to animate musical instruments. The tool can also animate 3D and anime characters.
Omnihuman struggles with laughter and expressive sounds other than talking or singing. Animating musical instruments is beyond its current capabilities. The AI correctly identifies multiple faces on the scene.
A review of Seaweed, Bite Dance's new video generator. It focuses on testing complex prompts like intense fights, gymnastic moves, breakdancing, and writing text and eating spaghetti. Seaweed excels in resolution and detail but lags in accurately following prompts and handling high-action scenes compared to models like Juan 2.1. It struggles with generating anime content and famous figures.
Omnihuman's lip-sync feature is the most impressive aspect of the Dreamina platform. The video encourages viewers to try the tools. Stay up to date with AI news via the newsletter.