Tiktok’s new IAs’ new IAs’ Animates’ photos and sound synchronized; Look at the examples

The Chinese company, the owner of Tiktok, announced the results of a new general artificial intelligence (AI). The platform is called Omnihuman-1, and according to the company, it can create extremely realistic videos for even a single photo.

Tongue model Produces short clips from a reference in the form of images or videos. Moreover, Synchronize the body and lip with a voice provided by the useras if it was initially a part of the material.

The preliminary results of Omnihuman-1 have been published by the developers team and viralized on social networks for high quality quality. In a few cases, It is even possible to realize that the content is produced by a number of inconsistencies by an AI.However, content has higher quality than other similar models.

The deep quilt of Tiktok’s owner

Videos created by Bandans AI are different from competitors, because they can represent animated people or numbers from the waist up or in the body, in several different poses that make sense according to content – for example, if the lesson or person is singing a song.

In addition, the results are considered more realistic, because the education itself is multi -moded: From the beginning, AI has been working with visual materials since training as an image set.The sound and movement that helps when changing or producing new content.

One of the examples described is a video of Albert Einstein as if he was giving a class or interview. The material used is a photo of the researcher and a pre -existing sound, giving life to the physical one for a moment. Attentive eyes and ears may notice the problems in automatic saess and the background of the frame in the background, but the result is already impressive and has the potential to improve.

https://www.youtube.com/watch?v=n6hkcs2pj0q

According to Bayta, Language Model Trained with people with 18,700 hours of video data. The company does not explain the source of these materials, which means that social networking materials and platforms such as Youtube are collected and feed on AI without authorizing it.

In another publication, the singer Taylor Swift, which provides a lot of participation, is in a show that plays songs in another language. The sound does not seem like it, but the synchronization work is remarkable. And as in all the other clip created with this AI, there is a large “water that defines the artificial writing of the content.

For now, omnihuman-1 It cannot be used for commercial use and still undergoing limited experiments and demonstrations. ByTedance has not provided details about when or when you will include the language model in the future, but technical details and more video samples can be controlled in the project warehouse.