AI tools are dominated these days. The biggest tech company is working on bringing a new AI tool. After Deepsek, now another Chinese company bytedance has launched the new AI Tool Omnihuman-1, which can make a full video from a photo. The company claims that it works better than the already existing model and can generate a very real looking video. In the video made from the photo, people are seen talking, using institute etc.
What is the special things of this AI tool?
In a research paper, the company has claimed that the new tool can make a video with every aspect ratio image. Whether it is a photo portrait, close-up or full body, this tool can generate a very real-looking high-quality video. The company says that this tool changes the input into stir, gestures and expressions with the help of multimodal conditioning. It is one step ahead of the model just changing facial expressions or adding human voice. In a video made from this tool on the company’s website, Albert Einstein is shown standing in front of a blackboard. In this, his hands are also seen moving.
Training is given from thousands of hours video
Tiktok owner by Teasedance said that videos of 18,700 hours of humans have been shown to train this tool. Apart from this, it has been trained with the help of text, physical poses and audio etc. It is believed that bytedance has given training to this tool with the help of videos on Tiktok. The company says that it makes a video-like video based on different inputs and it requires limited data.
Also read-
The iPhone will become the entire house remote! Apple bringing amazing technology, achieved patent
(tagstotranslate) deepseek
